The ViT from the base variant of the PE Core line of models.

Embeddings,PyTorch

The ViT from the large variant of the PE Core line of models.

Embeddings,PyTorch

Classic neural network that recognizes images and helped launch the deep learning revolution

Classification,Embeddings,Logits,Imagenet,PyTorch,Alexnet,Official

Fast single-image to 3D Gaussian splat model generating photorealistic novel views in under one second

3d,Gaussian-splatting,Novel-view,PyTorch,Transformer

Finds objects in high-resolution photos by pinpointing their centers with exceptional accuracy and speed

Detection,Coco,TensorFlow-2,Centernet

Efficient object finder optimized for medium-resolution images to run faster on regular computers

Detection,Coco,TensorFlow-2,Centernet

Lightweight object detector that runs smoothly on phones and other portable devices

Detection,Coco,TensorFlow-2,Centernet,Mobilenet

Advanced object finder with deeper processing for more accurate results in challenging scenes

Detection,Coco,TensorFlow-2,Centernet,Resnet

Balanced object detector that works well for most everyday computer vision tasks and applications

Detection,Coco,TensorFlow-2,Centernet,Resnet

Updated version with improved training stability for more consistent object detection across different images

Detection,Coco,TensorFlow-2,Centernet,Resnet

Vision transformer for image classification and custom fine-tuning on specialized datasets

Classification,Logits,Embeddings,PyTorch,Transformers,Official

Understands both images and text together, enabling search and classification using natural language descriptions

Classification,Logits,Embeddings,PyTorch,Clip,Zero-shot,Transformer,Official

Base modern CNN with transformer elements for robust visual understanding

Classification,Imagenet,PyTorch,Transformers,Convnext,Official,Embeddings

Large modern CNN demonstrating competitive performance with vision transformers

Classification,Imagenet,PyTorch,Transformers,Convnext,Embeddings,Official

Small modernized CNN delivering strong accuracy through architectural innovations

Classification,Imagenet,PyTorch,Transformers,Convnext,Official,Embeddings

Tiny modern CNN bridging traditional convolutions with transformer-inspired improvements

Classification,Imagenet,PyTorch,Transformers,Convnext,Embeddings,Official

Extra-large modern CNN maximizing architectural improvements for top accuracy

Classification,Imagenet,PyTorch,Transformers,Convnext,Official,Embeddings

Creates detailed pixel-by-pixel labels for urban scenes, helping autonomous vehicles understand their surroundings

Segmentation,Cityscapes,TensorFlow,Deeplabv3,Legacy

Efficient street scene labeler designed to run on phones and edge devices with limited resources

Segmentation,Cityscapes,TensorFlow,Deeplabv3,Legacy

Labels everyday objects in images pixel by pixel for general scene understanding and analysis

Segmentation,Coco,PyTorch,Resnet,Deeplabv3,Official

Faster version that quickly identifies and labels objects in images for real-time applications

Segmentation,Coco,PyTorch,Resnet,Deeplabv3,Official

Compact yet powerful classifier that delivers strong results while using minimal computational resources

Classification,Embeddings,Logits,Imagenet,PyTorch,Densenet,Official

Dense network that achieves high accuracy for image classification and adapts well to new tasks

Classification,Embeddings,Logits,Imagenet,PyTorch,Densenet,Official

Deeper variant offering improved accuracy while remaining efficient enough for practical deployment

Classification,Embeddings,Logits,Imagenet,PyTorch,Densenet,Official

Extra-deep model providing the most detailed features for complex image understanding tasks

Classification,Embeddings,Logits,Imagenet,PyTorch,Densenet,Official

Balanced Depth Anything V2 model for general-purpose monocular depth estimation

Depth,3d,PyTorch,Transformers

High-accuracy Depth Anything V2 model for detailed monocular depth estimation

Depth,3d,PyTorch,Transformers

Lightweight Depth Anything V2 model for fast monocular depth estimation

Depth,3d,PyTorch,Transformers

Balanced Depth Anything V3 model for general-purpose monocular depth estimation

Depth,3d,PyTorch,Transformer

Largest Depth Anything V3 model for maximum accuracy depth estimation

Depth,3d,PyTorch,Transformer

High-accuracy Depth Anything V3 model for detailed monocular depth estimation

Depth,3d,PyTorch,Transformer

Depth Anything V3 with metric depth output in meters and sky segmentation

Depth,3d,PyTorch,Transformer,Metric,Sky

Depth Anything V3 monocular model with relative depth and sky segmentation

Depth,3d,PyTorch,Transformer,Sky

Depth Anything V3 nested architecture combining giant encoder with large decoder

Depth,3d,PyTorch,Transformer

Lightweight Depth Anything V3 model for fast monocular depth estimation

Depth,3d,PyTorch,Transformer

Hugging Face Transformers model for monocular depth estimation

Depth,PyTorch,Transformers

Modern object detector that finds items in images without needing complex post-processing steps

Detection,Logits,Embeddings,PyTorch,Transformers,Official

D-FINE Large from "D-FINE: Redefine Regression Task in DETRs as Fine-grained Distribution Refinement" trained on COCO. Achieves 54.0% AP at 124 FPS on T4 GPU.

Detection,Coco,Embeddings,PyTorch,Transformers,Detr,Official

D-FINE Medium from "D-FINE: Redefine Regression Task in DETRs as Fine-grained Distribution Refinement" trained on COCO. Mid-size real-time object detector.

Detection,Coco,Embeddings,PyTorch,Transformers,Detr,Official

D-FINE Nano from "D-FINE: Redefine Regression Task in DETRs as Fine-grained Distribution Refinement" trained on COCO. Ultra-lightweight real-time object detector.

Detection,Coco,Embeddings,PyTorch,Transformers,Detr,Official

D-FINE Small from "D-FINE: Redefine Regression Task in DETRs as Fine-grained Distribution Refinement" trained on COCO. Balanced real-time object detector.

Detection,Coco,Embeddings,PyTorch,Transformers,Detr,Official

D-FINE XLarge from "D-FINE: Redefine Regression Task in DETRs as Fine-grained Distribution Refinement" trained on COCO. Achieves 55.8% AP at 78 FPS on T4 GPU.

Detection,Coco,Embeddings,PyTorch,Transformers,Detr,Official

Enhanced image search model that resists noise and errors for more reliable similarity matching

Embeddings,PyTorch,Dinov2,Transformer,Official

Creates searchable image fingerprints for finding similar pictures and organizing large photo collections

Embeddings,PyTorch,Dinov2,Transformer,Official

Highest-capacity dinov2 search model with maximum stability for finding images across massive diverse datasets

Embeddings,PyTorch,Dinov2,Transformer,Official

Powerful image search engine that handles enormous photo collections with rich detail extraction

Embeddings,PyTorch,Dinov2,Transformer,Official

Large stable model for finding and grouping similar images across big databases reliably

Embeddings,PyTorch,Dinov2,Transformer,Official

Large model that creates detailed image fingerprints for advanced search and automatic grouping

Embeddings,PyTorch,Dinov2,Transformer,Official

Compact stable model for image search that runs efficiently on phones and edge devices

Embeddings,PyTorch,Dinov2,Transformer,Official

Small model enabling image search and similarity matching directly on mobile devices

Embeddings,PyTorch,Dinov2,Transformer,Official

Tiny object detector optimized for phones and embedded systems working with smaller images

Detection,Coco,TensorFlow-2,Efficientdet

Legacy-compatible tiny object detector for older systems still running TensorFlow 1 frameworks

Detection,Coco,TensorFlow-1,Efficientdet,Legacy

Versatile object finder for medium-sized images supporting many different computer vision applications

Detection,Coco,TensorFlow-2,Efficientdet

Legacy version of versatile object finder maintaining compatibility with TensorFlow 1

Detection,Coco,TensorFlow-1,Efficientdet,Legacy

Balanced object detector offering good speed and accuracy for general-purpose image analysis

Detection,Coco,TensorFlow-2,Efficientdet

Legacy-compatible balanced detector for established pipelines still using TensorFlow 1

Detection,Coco,TensorFlow-1,Efficientdet,Legacy

Accurate object finder for larger images with better detection of big objects

Detection,Coco,TensorFlow-2,Efficientdet

Legacy object detector maintaining compatibility for systems using TensorFlow 1

Detection,Coco,TensorFlow-1,Efficientdet,Legacy

High-accuracy object detector for detailed images delivering precise results in complex scenes

Detection,Coco,TensorFlow-2,Efficientdet

Legacy high-accuracy object detector ensuring backward compatibility with TensorFlow 1

Detection,Coco,TensorFlow-1,Efficientdet,Legacy

Precision-focused object finder for very large images prioritizing accuracy over speed

Detection,Coco,TensorFlow-2,Efficientdet

Legacy object detector with the highest accuracy in its family using TensorFlow 1

Detection,Coco,TensorFlow-1,Efficientdet,Legacy

Deep object detector for large images achieving state-of-the-art accuracy on challenging content

Detection,Coco,TensorFlow-2,Efficientdet

Legacy deep detector maintaining top accuracy for mature TensorFlow 1 production stacks

Detection,Coco,TensorFlow-1,Efficientdet,Legacy

Maximum accuracy object finder for extra-large images pushing detection quality to the limit

Detection,Coco,TensorFlow-2,Efficientdet

Efficient image classifier optimized for mobile devices with excellent accuracy-efficiency tradeoff

Classification,Imagenet,PyTorch,Transformers,Efficientnet,Official,Embeddings

Scaled efficient classifier with improved accuracy for slightly larger computational budgets

Classification,Imagenet,PyTorch,Transformers,Efficientnet,Official,Embeddings

Balanced efficient model providing stronger performance while maintaining reasonable resource usage

Classification,Imagenet,PyTorch,Transformers,Efficientnet,Official,Embeddings

Mid-scale efficient classifier delivering high accuracy for versatile deployment scenarios

Classification,Imagenet,PyTorch,Transformers,Efficientnet,Official,Embeddings

Large efficient model with enhanced features for transfer learning applications

Classification,Imagenet,PyTorch,Transformers,Efficientnet,Official,Embeddings

High-capacity efficient classifier prioritizing accuracy with available compute resources

Classification,Imagenet,PyTorch,Transformers,Efficientnet,Official,Embeddings

Extended efficient model approaching state-of-the-art accuracy on challenging datasets

Classification,Imagenet,PyTorch,Transformers,Efficientnet,Official,Embeddings

Maximum efficient classifier pushing performance boundaries while preserving efficiency principles

Classification,Imagenet,PyTorch,Transformers,Efficientnet,Official,Embeddings

High-accuracy object finder that sees wider context for better detection in complex scenes

Detection,Coco,TensorFlow,Faster-rcnn,Inception,Resnet

Speed-optimized detector that runs faster by examining fewer regions while maintaining good accuracy

Detection,Coco,TensorFlow,Faster-rcnn,Inception,Resnet,Legacy

Compact object detector achieving real-time speeds for responsive computer vision applications

Detection,Coco,TensorFlow,Faster-rcnn,Inception

Smart detector using NAS-designed architecture for improved object finding across diverse images

Detection,Coco,TensorFlow,Faster-rcnn

Fast NAS-designed detector that speeds up processing for time-sensitive applications and deployments

Detection,Coco,TensorFlow,Faster-rcnn,Legacy

Deep object detector balancing accuracy and speed for reliable performance across varied scenes

Detection,Coco,TensorFlow,Faster-rcnn,Resnet,Legacy

Accelerated deep detector that processes fewer regions for faster results with minimal accuracy loss

Detection,Coco,TensorFlow,Faster-rcnn,Resnet,Legacy

Versatile object detector suitable for everyday vision tasks in research and production environments

Detection,Coco,TensorFlow,Faster-rcnn,Resnet

Multi-scale object finder that accurately detects both small and large items in images

Detection,Coco,PyTorch,Faster-rcnn,Resnet,Official

Speed-focused detector optimized for running on embedded devices and resource-limited hardware

Detection,Coco,TensorFlow,Faster-rcnn,Resnet,Legacy

Open-vocabulary instance segmentation with FC-CLIP (COCO); returns only "thing" segments as instance masks. Uses frozen ConvNeXt-Large CLIP backbone + Mask2Former decoder.

Instances,Zero-shot,PyTorch,Transformers,Official

Open-vocabulary panoptic segmentation with FC-CLIP (COCO); returns both thing and stuff segments as instance masks. Uses frozen ConvNeXt-Large CLIP backbone + Mask2Former decoder.

Panoptic,Zero-shot,PyTorch,Transformers,Official

Creates detailed pixel-level labels for images, identifying and outlining twenty-one different object categories

Segmentation,Coco,PyTorch,Fcn,Resnet,Official

Fast image labeler that quickly identifies and outlines objects for interactive editing and annotation

Segmentation,Coco,PyTorch,Fcn,Resnet,Official

Classic image classifier providing reliable categorization and features for various computer vision projects.

Classification,Embeddings,Logits,Imagenet,PyTorch,Googlenet,Official

Full-size open-set object detector using a Swin-B vision backbone and BERT text encoder, with higher accuracy across a wide range of vocabulary

Detection,PyTorch,Transformers,Zero-shot,Official

Compact open-set object detector that finds objects described in natural language, using a Swin-T vision backbone and BERT text encoder

Detection,PyTorch,Transformers,Zero-shot,Official

Hugging Face Transformers model for zero-shot semantic segmentation

Segmentation,Embeddings,PyTorch,Transformers,Zero-shot,Official

High-accuracy image classifier with advanced architecture for precise categorization and feature extraction

Classification,Embeddings,Logits,Imagenet,TensorFlow-1,Inception,Resnet

Efficient image classifier delivering accurate results with useful features for transfer learning applications

Classification,Embeddings,Logits,Imagenet,PyTorch,Inception,Official

Enhanced image classifier with deeper architecture improving accuracy for demanding vision tasks

Classification,Embeddings,Logits,Imagenet,TensorFlow-1,Inception,Legacy

Finds people in images and maps their body joints for pose estimation and motion analysis

Keypoints,Coco,PyTorch,Keypoint-rcnn,Resnet,Official

Balanced open-vocabulary detector that recognizes objects from natural-language prompts with strong accuracy.

Detection,PyTorch,Transformers,Zero-shot,Official

High-capacity open-vocabulary detector for maximum detection quality from free-form text queries.

Detection,PyTorch,Transformers,Zero-shot,Official

Lightweight open-vocabulary detector that finds any object you describe in images without class-specific training.

Detection,PyTorch,Transformers,Zero-shot,Official

Creates precise object outlines and boxes for detailed scene understanding in high-resolution images

Instances,Coco,TensorFlow,Mask-rcnn,Inception,Resnet

Fast object outliner generating masks and boxes with lower computing requirements for real-time use

Instances,Coco,TensorFlow,Mask-rcnn,Inception

Enhanced object outliner providing detailed masks with better handling of large objects in scenes

Instances,Coco,TensorFlow,Mask-rcnn,Resnet,Legacy

General-purpose object outliner creating masks and boxes suitable for most vision analysis tasks

Instances,Coco,TensorFlow,Mask-rcnn,Resnet,Legacy

Multi-scale object outliner using advanced architecture for accurate segmentation across different object sizes

Instances,Coco,PyTorch,Mask-rcnn,Resnet,Official

Mask2Former with Swin-B backbone for semantic segmentation on ADE20K (150 classes)

Segmentation,PyTorch,Transformers,Official

Mask2Former with Swin-B backbone for instance segmentation on COCO, returning per-instance binary masks

Instances,PyTorch,Transformers,Official

Mask2Former with Swin-B backbone for panoptic segmentation on COCO, unifying things and stuff into a single segmentation

Panoptic,PyTorch,Transformers,Official

Mask2Former with Swin-L backbone for semantic segmentation on ADE20K (150 classes)

Segmentation,PyTorch,Transformers,Official

Mask2Former with Swin-L backbone for instance segmentation on COCO, returning per-instance binary masks

Instances,PyTorch,Transformers,Official

Mask2Former with Swin-L backbone for panoptic segmentation on COCO, unifying things and stuff into a single segmentation

Panoptic,PyTorch,Transformers,Official

Mask2Former with Swin-S backbone for semantic segmentation on ADE20K (150 classes)

Segmentation,PyTorch,Transformers,Official

Mask2Former with Swin-S backbone for instance segmentation on COCO, returning per-instance binary masks

Instances,PyTorch,Transformers,Official

Mask2Former with Swin-S backbone for panoptic segmentation on COCO, unifying things and stuff into a single segmentation

Panoptic,PyTorch,Transformers,Official

Mask2Former with Swin-T backbone for semantic segmentation on ADE20K (150 classes)

Segmentation,PyTorch,Transformers,Official

Mask2Former with Swin-T backbone for instance segmentation on COCO, returning per-instance binary masks

Instances,PyTorch,Transformers,Official

Mask2Former with Swin-T backbone for panoptic segmentation on COCO, unifying things and stuff into a single segmentation

Panoptic,PyTorch,Transformers,Official

Medical segmentation tool that outlines organs and structures in medical videos and 3D scans

Segment-anything,PyTorch,Zero-shot,Video,Med-sam,Transformer,Official

Medical SigLIP for zero-shot image classification and embeddings

Classification,PyTorch,Official,Medical,Zero-shot,Embeddings

Ultra-lightweight image classifier designed by AI for running directly on phones and IoT devices

Classification,Embeddings,Logits,Imagenet,PyTorch,Mnasnet,Official

Mobile-optimized classifier balancing size and accuracy for efficient on-device image recognition

Classification,Embeddings,Logits,Imagenet,PyTorch,Mnasnet,Official

Efficient mobile classifier using specialized architecture for fast image recognition on phones

Classification,Embeddings,Logits,Imagenet,TensorFlow-1,Mobilenet

Mobile-friendly image classifier optimized for quick training and deployment on resource-limited devices

Classification,Embeddings,Logits,Imagenet,PyTorch,Mobilenet,Official

CLIP‑based vision‑language model for zero‑shot dermatology image classification.

Classification,PyTorch,Official,Medical,Embeddings

Real-time detector that finds any object you describe in words, perfect for live video analysis

Detection,Logits,Embeddings,PyTorch,Transformers,Zero-shot,Official

OneFormer with Swin-L backbone trained on ADE20K for instance segmentation, returning per-instance binary masks

Instances,PyTorch,Transformers,Official

OneFormer with Swin-L backbone trained on ADE20K for panoptic segmentation, returning per-segment binary masks

Panoptic,PyTorch,Transformers,Official

OneFormer with Swin-L backbone trained on ADE20K for semantic segmentation

Segmentation,PyTorch,Transformers,Official

OneFormer with Swin-T backbone trained on ADE20K for instance segmentation, returning per-instance binary masks

Instances,PyTorch,Transformers,Official

OneFormer with Swin-T backbone trained on ADE20K for panoptic segmentation, returning per-segment binary masks

Panoptic,PyTorch,Transformers,Official

OneFormer with Swin-T backbone trained on ADE20K for semantic segmentation

Segmentation,PyTorch,Transformers,Official

OneFormer with Swin-L backbone trained on Cityscapes for instance segmentation, returning per-instance binary masks

Instances,PyTorch,Transformers,Official

OneFormer with Swin-L backbone trained on Cityscapes for panoptic segmentation, returning per-segment binary masks

Panoptic,PyTorch,Transformers,Official

OneFormer with Swin-L backbone trained on Cityscapes for semantic segmentation

Segmentation,PyTorch,Transformers,Official

OneFormer with Swin-L backbone trained on MS COCO 2017 for instance segmentation, returning per-instance binary masks

Instances,PyTorch,Transformers,Official

OneFormer with Swin-L backbone trained on MS COCO 2017 for panoptic segmentation, returning per-segment binary masks

Panoptic,PyTorch,Transformers,Official

OneFormer with Swin-L backbone trained on MS COCO 2017 for semantic segmentation

Segmentation,PyTorch,Transformers,Official

Connects images with text descriptions enabling search by words and automatic content filtering systems

Classification,Logits,Embeddings,PyTorch,Clip,Zero-shot,Transformer

Finds any object you name in pictures using 16x16 image patches without needing specific training for those items

Detection,Logits,PyTorch,Transformers,Zero-shot,Official

Finds any object you name in pictures using efficient 32x32 image patches without needing specific training

Detection,Logits,PyTorch,Transformers,Zero-shot,Official

Large OWL-ViT zero-shot object detector with ViT-L/14 backbone. Achieves higher accuracy than base models, especially for smaller objects.

Detection,Logits,PyTorch,Transformers,Zero-shot,Official

PP-OCRv6 medium text detection model that locates text regions in images

Detection,Ocr,PyTorch

PP-OCRv6 medium OCR model that detects and reads text in images

Ocr,Detection,PyTorch

PP-OCRv6 small text detection model that locates text regions in images

Detection,Ocr,PyTorch

PP-OCRv6 small OCR model that detects and reads text in images

Ocr,Detection,PyTorch

PP-OCRv6 tiny text detection model that locates text regions in images

Detection,Ocr,PyTorch

PP-OCRv6 tiny OCR model that detects and reads text in images

Ocr,Detection,PyTorch

Vision Transformer for pose estimation with 90M parameters removing complex decoder components.

Keypoints,Coco,PyTorch,Transformers,Pose-estimation

Zero-shot medical image classifier trained on biomedical image–text pairs.

Classification,PyTorch,Official,Medical,Zero-shot,Embeddings

Compact vision-language model for object detection via 2D grounding

Detection,Vlm,PyTorch,Transformer,Zero-shot

Balanced vision-language model for object detection via 2D grounding

Detection,Vlm,PyTorch,Transformer,Zero-shot

High-quality vision-language model for object detection via 2D grounding

Detection,Vlm,PyTorch,Transformer,Zero-shot

Multimodal embedding model for image-text similarity and retrieval

Embeddings,Vlm,PyTorch,Transformer

High-quality multimodal embedding model for image-text similarity and retrieval

Embeddings,Vlm,PyTorch,Transformer

Classic image recognition model providing reliable categorization and visual features for many applications

Classification,Embeddings,Logits,Imagenet,TensorFlow-1,Resnet,Legacy

Improved image classifier with smoother training process and better features for adapting to new tasks

Classification,Embeddings,Logits,Imagenet,TensorFlow-1,Resnet,Legacy

Deep image recognition model delivering high accuracy for demanding classification and analysis tasks

Classification,Embeddings,Logits,Imagenet,PyTorch,Resnet,Official

Very deep classifier providing the richest visual features for precision-critical image understanding applications

Classification,Embeddings,Logits,Imagenet,PyTorch,Resnet,Official

Lightweight image classifier designed for fast recognition on phones and other resource-limited devices

Classification,Embeddings,Logits,Imagenet,PyTorch,Resnet,Official

Balanced image classifier offering good accuracy and speed for everyday computer vision needs

Classification,Embeddings,Logits,Imagenet,PyTorch,Resnet,Official

Most popular image recognition backbone widely used as starting point for custom vision projects

Classification,Embeddings,Logits,Imagenet,PyTorch,Resnet,Official

Powerful image classifier with enhanced capacity for handling complex visual recognition challenges effectively

Classification,Embeddings,Logits,Imagenet,PyTorch,Resnext,Official

Efficient advanced classifier delivering strong accuracy with reasonable computing requirements for practical deployments

Classification,Embeddings,Logits,Imagenet,PyTorch,Resnext,Official

Fast object detector that quickly finds and boxes eighty common items in any image

Detection,Coco,PyTorch,Retinanet,Resnet,Official

Efficient object finder producing accurate boxes for eighty object types with optimized processing speed

Detection,Coco,TensorFlow,Rfcn,Resnet,Legacy

RF-DETR Base real-time object detector trained on COCO. High accuracy for general use.

Detection,Coco,PyTorch,Transformer,Rfdetr,Official

RF-DETR Large real-time object detector trained on COCO. Maximum detection accuracy.

Detection,Coco,PyTorch,Transformer,Rfdetr,Official

RF-DETR Medium real-time object detector trained on COCO. Balanced speed and accuracy.

Detection,Coco,PyTorch,Transformer,Rfdetr,Official

RF-DETR Nano real-time object detector trained on COCO. Ultra-fast, suitable for edge devices.

Detection,Coco,PyTorch,Transformer,Rfdetr,Official

RF-DETR Seg 2XLarge instance segmentation model trained on COCO. Maximum segmentation accuracy.

Instances,Coco,PyTorch,Transformer,Rfdetr,Official

RF-DETR Seg Large instance segmentation model trained on COCO. High accuracy segmentation.

Instances,Coco,PyTorch,Transformer,Rfdetr,Official

RF-DETR Seg Medium instance segmentation model trained on COCO. Balanced speed and accuracy.

Instances,Coco,PyTorch,Transformer,Rfdetr,Official

RF-DETR Seg Nano instance segmentation model trained on COCO. Ultra-fast, suitable for edge devices.

Instances,Coco,PyTorch,Transformer,Rfdetr,Official

RF-DETR Seg Small instance segmentation model trained on COCO. Fast with good accuracy.

Instances,Coco,PyTorch,Transformer,Rfdetr,Official

RF-DETR Seg XLarge instance segmentation model trained on COCO. Very high accuracy.

Instances,Coco,PyTorch,Transformer,Rfdetr,Official

RF-DETR Small real-time object detector trained on COCO. Fast with good accuracy.

Detection,Coco,PyTorch,Transformer,Rfdetr,Official

Modern real-time object detector that finds items without complex post-processing for responsive applications

Detection,Coco,PyTorch,Transformer,Rtdetr,Official

High-accuracy real-time object detector with ResNet-101 backbone, offering the best accuracy in the RT-DETRv2 family

Detection,Coco,Embeddings,PyTorch,Transformers,Rtdetr,Official

Balanced real-time object detector offering improved accuracy for production use

Detection,Coco,Embeddings,PyTorch,Transformers,Rtdetr,Official

Lightweight real-time object detector optimized for speed on edge devices

Detection,Coco,Embeddings,PyTorch,Transformers,Rtdetr,Official

High-capacity object detector delivering very precise results at speeds suitable for production use

Detection,Coco,PyTorch,Transformer,Rtdetr,Official

Efficient transformer-based semantic segmentation model for scene parsing with 150 classes

Segmentation,PyTorch,Segformer,Official,Embeddings

Balanced SegFormer model providing good accuracy-efficiency tradeoff for scene understanding

Segmentation,PyTorch,Segformer,Official,Embeddings

Medium-sized SegFormer delivering enhanced segmentation quality for complex scenes

Segmentation,PyTorch,Segformer,Official,Embeddings

Larger SegFormer model with improved accuracy for detailed semantic segmentation

Segmentation,PyTorch,Segformer,Official,Embeddings

High-capacity SegFormer achieving excellent results on challenging segmentation tasks

Segmentation,PyTorch,Segformer,Official,Embeddings

Largest SegFormer model delivering the best semantic segmentation performance in its family

Segmentation,PyTorch,Segformer,Official,Embeddings

Accurate image segmentation model for editing, labeling, and creative work with still pictures

Segment-anything,PyTorch,Zero-shot,Transformer,Official

Video segmentation model that tracks and outlines objects throughout clips for editing and analysis

Segment-anything,PyTorch,Zero-shot,Video,Transformer,Official

High-quality image segmenter producing detailed masks for demanding professional editing and annotation tasks

Segment-anything,PyTorch,Zero-shot,Transformer,Official

Advanced video segmenter providing fine object tracking throughout full videos for post-production work

Segment-anything,PyTorch,Zero-shot,Video,Transformer,Official

Fast image segmentation model that runs efficiently on laptops and edge computing devices

Segment-anything,PyTorch,Zero-shot,Transformer,Official

Quick video segmentation model delivering rapid object tracking on standard graphics cards

Segment-anything,PyTorch,Zero-shot,Video,Transformer,Official

Smallest image segmentation model offering instant results for mobile apps and embedded systems

Segment-anything,PyTorch,Zero-shot,Official

Tiny video segmenter enabling real-time object tracking on phones and compact devices

Segment-anything,PyTorch,Zero-shot,Video,Transformer,Official

Updated image segmenter with improved mask accuracy for everyday editing and dataset creation

Segment-anything,PyTorch,Zero-shot,Transformer,Official

Enhanced video segmenter with better tracking quality for video analysis and scene understanding

Segment-anything,PyTorch,Zero-shot,Video,Transformer,Official

Large updated model offering even finer masks for high-resolution professional image workflows

Segment-anything,PyTorch,Zero-shot,Transformer,Official

Large video model producing exceptionally detailed masks throughout long videos for intensive production

Segment-anything,PyTorch,Zero-shot,Video,Transformer,Official

Balanced updated segmenter combining speed and accuracy for edge device image processing

Segment-anything,PyTorch,Zero-shot,Transformer,Official

Improved video segmenter maintaining quick performance on compact hardware while enhancing mask quality

Segment-anything,PyTorch,Zero-shot,Video,Transformer,Official

Enhanced mobile image segmenter for apps, augmented reality filters, and on-device processing

Segment-anything,PyTorch,Zero-shot,Transformer,Official

Upgraded mobile video segmenter for live effects on phones, wearables, and smart cameras

Segment-anything,PyTorch,Zero-shot,Video,Transformer,Official

Open-vocabulary instance segmentation that finds and segments all objects matching a text concept like 'person' or 'yellow school bus'

Segment-anything,PyTorch,Zero-shot,Transformer,Official

Open-vocabulary video segmentation that finds, segments, and tracks all objects matching a text concept across video frames

Segment-anything,PyTorch,Zero-shot,Video,Transformer,Official

Interactive segmentation tool that instantly outlines any object you point to or describe

Segment-anything,Sa-1b,PyTorch,Zero-shot,Transformer,Official

Highest quality segmentation model creating extremely detailed masks for research and large-scale annotation projects

Segment-anything,Sa-1b,PyTorch,Zero-shot,Transformer,Official

Large segmentation model producing finer object outlines for professional editing and labeling workflows

Segment-anything,Sa-1b,PyTorch,Zero-shot,Transformer,Official

Hugging Face Transformers model for semantic segmentation

Segmentation,PyTorch,Transformers,Official,Embeddings

Ultra-small image classifier for tiny devices with very limited power and memory

Classification,Embeddings,Logits,Imagenet,PyTorch,Shufflenet,Official

Mobile image classifier that works efficiently on phones with modest computing resources

Classification,Embeddings,Logits,Imagenet,PyTorch,Shufflenet,Official

Hugging Face Transformers model for zero-shot image classification

Classification,Logits,Embeddings,PyTorch,Transformers,Zero-shot,Official

Tiny image classifier that fits in just five megabytes for embedded devices

Classification,Imagenet,PyTorch,Squeezenet,Official

Ultra-compact image classifier perfect for severely resource-constrained hardware and applications

Classification,Imagenet,PyTorch,Squeezenet,Official

Real-time object finder that quickly identifies eighty common items in any image

Detection,Coco,TensorFlow,Ssd,Inception

Mobile object detector that runs smoothly on phones and edge computing processors

Detection,Coco,TensorFlow,Ssd,Mobilenet,Legacy

Enhanced mobile detector that better finds small objects in larger resolution images

Detection,Coco,TensorFlow-2,Ssd,Mobilenet

Mobile object detector that runs smoothly on phones and edge computing processors

Detection,Coco,TensorFlow,Ssd,Mobilenet,Legacy

Fast object finder optimized for quick GPU processing of smaller input images

Detection,Coco,TensorFlow-2,Ssd,Mobilenet

Accurate object detector combining strong backbone with multi-scale detection for better results

Detection,Coco,TensorFlow,Ssd,Resnet,Legacy

Base hierarchical transformer delivering strong results across vision tasks

Classification,Imagenet,PyTorch,Transformers,Swin-transformer,Official,Embeddings

Large hierarchical transformer with enhanced capacity for demanding applications

Classification,Imagenet,PyTorch,Transformers,Swin-transformer,Official,Embeddings

Small hierarchical transformer balancing efficiency and performance for practical use

Classification,Imagenet,PyTorch,Transformers,Swin-transformer,Official,Embeddings

Tiny hierarchical transformer for efficient visual recognition on edge devices

Classification,Imagenet,PyTorch,Transformers,Swin-transformer,Official,Embeddings

An enhanced multimodal embedding model that extends the capabilities of Marengo 2.7 with support for text and image interleaved input modality. This model generates high-quality vector representations of video, text, audio, image, and interleaved text-image content for similarity search, clustering, and other machine learning tasks

Video,Embeddings,Official

A multimodal model that provides comprehensive video understanding and analysis capabilities, including content recognition, scene detection, and contextual understanding

Video,Caption,Official

A multimodal model that provides comprehensive video understanding and analysis capabilities, including video-to-text generation, structured video segmentation, image-guided prompting, and timestamp-level temporal grounding

Video,Caption,Official

Hugging Face Transformers model for universal segmentation (instance, panoptic, or semantic)

Instances,Panoptic,Segmentation,PyTorch,Transformers,Official

Classic image classifier with stable training useful for various computer vision projects

Classification,Embeddings,Logits,Imagenet,PyTorch,Vgg,Official

Simple baseline image classifier valuable for research experimentation and learning purposes

Classification,Embeddings,Logits,Imagenet,PyTorch,Vgg,Official

Deeper classic classifier providing stable training process and solid accuracy results overall

Classification,Embeddings,Logits,Imagenet,PyTorch,Vgg,Official

Straightforward image classifier valued for easy experimentation and model compression studies

Classification,Embeddings,Logits,Imagenet,PyTorch,Vgg,Official

Popular feature extractor widely used for detection, style transfer, and medical imaging

Classification,Embeddings,Logits,Imagenet,PyTorch,Vgg,Official

TensorFlow version of the classic image classifier supporting legacy production systems

Classification,Embeddings,Logits,Imagenet,TensorFlow-1,Vgg,Legacy

PyTorch version of the popular classifier ready for modern deep learning workflows

Classification,Embeddings,Logits,Imagenet,PyTorch,Vgg,Official

Deep classic model providing rich features for style transfer and interpretability analysis

Classification,Embeddings,Logits,Imagenet,PyTorch,Vgg,Official

Deep image classifier delivering detailed features for creative applications and research projects

Classification,Embeddings,Logits,Imagenet,PyTorch,Vgg,Official

Modern image classifier that recognizes objects and provides useful features for various computer vision tasks

Classification,Logits,Embeddings,PyTorch,Transformers,Official

Simplified ViTPose with 90M parameters removing complex decoder components. Maintains 75.1 AP through direct heatmap prediction from transformer features. Streamlined architecture for easier deployment while preserving accuracy on human pose tasks.

Keypoints,Coco,PyTorch,Transformers,Pose-estimation,Official

Vision Transformer for pose estimation with 90M parameters using standard ViT backbone. Detects 17 human keypoints through heatmap regression achieving 75.8 AP on COCO. Processes 256x192 images with hierarchical features for accurate joint localization.

Keypoints,Coco,PyTorch,Transformers,Pose-estimation,Official

Base ViTPose+ with 130M parameters implementing mixture-of-experts modules. Delivers 77.5 AP through dataset-aware routing. MOE design handles multiple pose datasets simultaneously while maintaining strong per-dataset performance.

Keypoints,Coco,PyTorch,Transformers,Pose-estimation,Official

Huge ViTPose+ with 900M parameters maximizing MOE capacity for best performance. Delivers 78.9 AP through massive scale. Flagship mixture-of-experts model handling diverse pose datasets with dataset-specific optimization paths.

Keypoints,Coco,PyTorch,Transformers,Pose-estimation,Official

Large ViTPose+ with 430M parameters scaling MOE architecture for superior accuracy. Achieves 78.3 AP on COCO through enhanced capacity. Mixture-of-experts enables specialization across pose datasets while maintaining unified architecture.

Keypoints,Coco,PyTorch,Transformers,Pose-estimation,Official

Small ViTPose+ with 30M parameters using mixture-of-experts for multi-dataset training. Achieves 68.7 AP through dataset-specific adaptation. Lightweight MOE architecture enables efficient pose estimation across diverse human pose datasets.

Keypoints,Coco,PyTorch,Transformers,Pose-estimation,Official

Extra-wide deep classifier for high-precision image recognition and advanced transfer learning tasks

Classification,Embeddings,Logits,Imagenet,PyTorch,Wide-resnet,Official

Wide classifier offering stronger accuracy and better features for adapting to new tasks

Classification,Embeddings,Logits,Imagenet,PyTorch,Wide-resnet,Official

AI-designed detector family offering three model variants for diverse deployment scenarios

Detection,PyTorch,Yolo,Official

Classic real-time detector finding eighty object types quickly for video analysis applications

Detection,Coco,TensorFlow-1,Yolo,Legacy

Real-time object detector balancing high accuracy with fast processing speeds effectively

Detection,Coco,PyTorch,Yolo,Official

Model creating detailed object outlines for precise image editing and analysis

Instances,Coco,PyTorch,Yolo,Official

Object detector offering good balance between speed and accuracy for most applications

Detection,Coco,PyTorch,Yolo,Official

Model generating object masks efficiently for everyday segmentation tasks

Instances,Coco,PyTorch,Yolo,Official

Object detector designed specifically for phones and other edge computing devices

Detection,Coco,PyTorch,Yolo,Official

Edge model producing object outlines directly on phones and edge devices

Instances,Coco,PyTorch,Yolo,Official

Fast object detector ideal for systems with limited graphics processing power

Detection,Coco,PyTorch,Yolo,Official

Model creating object masks quickly for real-time segmentation applications

Instances,Coco,PyTorch,Yolo,Official

Object detector prioritizing accuracy over processing speed for critical applications

Detection,Coco,PyTorch,Yolo,Official

Model delivering high-quality object outlines for professional workflows

Instances,Coco,PyTorch,Yolo,Official

High-accuracy image classifier

Classification,Imagenet,PyTorch,Yolo,Official

High-accuracy NMS-free detector for demanding tasks

Detection,Coco,PyTorch,Yolo,Official

High-accuracy NMS-free pose estimation

Keypoints,Coco,PyTorch,Yolo,Official

High-accuracy NMS-free instance segmentation

Instances,Coco,PyTorch,Yolo,Official

Medium image classifier for general use

Classification,Imagenet,PyTorch,Yolo,Official

NMS-free detector with improved small object detection

Detection,Coco,PyTorch,Yolo,Official

NMS-free pose estimation with improved keypoint accuracy

Keypoints,Coco,PyTorch,Yolo,Official

NMS-free instance segmentation with improved masks

Instances,Coco,PyTorch,Yolo,Official

Ultra-fast image classifier for edge devices

Classification,Imagenet,PyTorch,Yolo,Official

Ultra-fast NMS-free detector for edge devices

Detection,Coco,PyTorch,Yolo,Official

Ultra-fast NMS-free pose estimation for edge devices

Keypoints,Coco,PyTorch,Yolo,Official

Ultra-fast NMS-free instance segmentation

Instances,Coco,PyTorch,Yolo,Official

Fast image classifier balancing speed and accuracy

Classification,Imagenet,PyTorch,Yolo,Official

Fast NMS-free detector balancing speed and accuracy

Detection,Coco,PyTorch,Yolo,Official

Fast NMS-free pose estimation balancing speed and accuracy

Keypoints,Coco,PyTorch,Yolo,Official

Fast NMS-free instance segmentation

Instances,Coco,PyTorch,Yolo,Official

Maximum accuracy image classifier

Classification,Imagenet,PyTorch,Yolo,Official

Maximum accuracy NMS-free detector

Detection,Coco,PyTorch,Yolo,Official

Maximum accuracy NMS-free pose estimation

Keypoints,Coco,PyTorch,Yolo,Official

Maximum accuracy NMS-free instance segmentation

Instances,Coco,PyTorch,Yolo,Official

Real-time model creating both object outlines and boxes for any described item

Instances,PyTorch,Yolo,Zero-shot,Official

Model producing masks and boxes for objects described in natural language

Instances,PyTorch,Yolo,Zero-shot,Official

Segments specified classes, generating object outlines and boxes for real-time applications

Instances,PyTorch,Yolo,Zero-shot,Official

Prompt-free large segmentation with 4,585 built-in classes

Instances,PyTorch,Yolo,Zero-shot,Prompt-free,Official

Open-vocabulary large segmentation with text/visual prompts

Instances,PyTorch,Yolo,Zero-shot,Official

Prompt-free medium segmentation with 4,585 built-in classes

Instances,PyTorch,Yolo,Zero-shot,Prompt-free,Official

Open-vocabulary medium segmentation with text/visual prompts

Instances,PyTorch,Yolo,Zero-shot,Official

Prompt-free nano segmentation with 4,585 built-in classes

Instances,PyTorch,Yolo,Zero-shot,Prompt-free,Official

Open-vocabulary nano segmentation with text/visual prompts

Instances,PyTorch,Yolo,Zero-shot,Official

Prompt-free small segmentation with 4,585 built-in classes

Instances,PyTorch,Yolo,Zero-shot,Prompt-free,Official

Open-vocabulary small segmentation with text/visual prompts

Instances,PyTorch,Yolo,Zero-shot,Official

Prompt-free extra-large segmentation with 4,585 built-in classes

Instances,PyTorch,Yolo,Zero-shot,Prompt-free,Official

Open-vocabulary extra-large segmentation with text/visual prompts

Instances,PyTorch,Yolo,Zero-shot,Official

Model outlining and boxing any object you describe without specific training

Instances,PyTorch,Yolo,Zero-shot,Official

Model creating masks for any object type you name in text

Instances,PyTorch,Yolo,Zero-shot,Official

Compact model producing outlines for objects described in words on edge devices

Instances,PyTorch,Yolo,Zero-shot,Official

Object detector with special optimizations for even faster inference on modern hardware

Detection,Coco,PyTorch,Yolo,Official

Balanced detector providing good accuracy and speed for general-purpose object detection tasks

Detection,Coco,PyTorch,Yolo,Official

Edge-optimized detector for devices with minimal computing resources available

Detection,Coco,PyTorch,Yolo,Official

Fast lightweight detector suitable for systems with limited GPU capabilities and memory

Detection,Coco,PyTorch,Yolo,Official

High-accuracy detector for demanding object detection applications and research

Detection,Coco,PyTorch,Yolo,Official

Real-time detector producing accurate results quickly for demanding vision applications

Detection,Coco,PyTorch,Yolo,Official

Real-time detector balancing good accuracy with fast processing speeds

Detection,Coco,PyTorch,Yolo,Official

Lightweight detector for edge devices needing basic object detection capabilities

Detection,Coco,PyTorch,Yolo,Official

Real-time detector delivering good results with minimal computational requirements

Detection,Coco,PyTorch,Yolo,Official

High-accuracy detector offering top precision for applications where quality is critical

Detection,Coco,PyTorch,Yolo,Official

Real-time detector with advanced architecture for improved object finding in complex scenes

Detection,Coco,PyTorch,Yolo,Official

Specialized detector that finds rotated objects in aerial and satellite imagery accurately

Detection,PyTorch,Yolo,Polylines,Obb,Official

General-purpose detector trained on diverse images recognizing over six hundred object categories

Detection,Oiv7,PyTorch,Yolo,Official

Creates precise object outlines for detailed image editing and analysis tasks

Instances,Coco,PyTorch,Yolo,Official

Finds and boxes any object you describe using natural language prompts

Detection,PyTorch,Yolo,Zero-shot,Official

Detector balancing speed and accuracy for everyday object detection needs

Detection,Coco,PyTorch,Yolo,Official

Finds rotated bounding boxes in aerial images for mapping and surveillance applications

Detection,PyTorch,Yolo,Polylines,Obb,Official

Versatile detector recognizing hundreds of different object types across varied image domains

Detection,Oiv7,PyTorch,Yolo,Official

Generates object masks with good balance of speed and quality

Instances,Coco,PyTorch,Yolo,Official

Detector understanding text descriptions to find matching objects in images

Detection,PyTorch,Yolo,Zero-shot,Official

Edge-optimized detector recognizing common objects on resource-limited devices effectively

Detection,Coco,PyTorch,Yolo,Official

Lightweight detector for finding rotated objects in aerial imagery on edge hardware

Detection,PyTorch,Yolo,Polylines,Obb,Official

Edge-friendly detector recognizing hundreds of object categories on resource-limited devices effectively

Detection,Oiv7,PyTorch,Yolo,Official

Edge-optimized model producing object outlines on devices with limited resources.

Instances,Coco,PyTorch,Yolo,Official

Detector offering fast performance on mid-range graphics cards and processors

Detection,Coco,PyTorch,Yolo,Official

Efficiently finds rotated objects in aerial photos for mapping and analysis tasks

Detection,PyTorch,Yolo,Polylines,Obb,Official

Compact detector recognizing diverse object types across many different image categories

Detection,Oiv7,PyTorch,Yolo,Official

Fast model creating object masks for real-time image segmentation needs

Instances,Coco,PyTorch,Yolo,Official

Lightweight detector finding objects based on text descriptions for edge applications

Detection,PyTorch,Yolo,Zero-shot,Official

High-accuracy detector for critical applications where precision matters most

Detection,Coco,PyTorch,Yolo,Official

High-precision detector for rotated objects in aerial and satellite imagery analysis

Detection,PyTorch,Yolo,Polylines,Obb,Official

Accurate general detector recognizing over six hundred different object types

Detection,Oiv7,PyTorch,Yolo,Official

High-accuracy model generating detailed object outlines for demanding professional applications

Instances,Coco,PyTorch,Yolo,Official

Open-vocabulary detector with high accuracy for text-based object finding

Detection,PyTorch,Yolo,Zero-shot,Official

Detector enhanced with transformer technology for improved object finding capabilities

Detection,Coco,PyTorch,Yolo,Official

Compact model producing both masks and boxes with transformer-enhanced accuracy

Instances,Coco,PyTorch,Yolo,Official

Advanced detector with transformer backbone delivering superior accuracy for complex scenes

Detection,Coco,PyTorch,Yolo,Official

Advanced model creating precise object outlines using enhanced transformer architecture

Instances,Coco,PyTorch,Yolo,Official

Finds any object you name in images without requiring training on those specific items

Classification,Logits,Embeddings,PyTorch,Transformers,Zero-shot,Official

Hugging Face Transformers model for zero-shot object detection

Detection,Logits,PyTorch,Transformers,Zero-shot,Official

NVIDIA LocateAnything-3B\\: open-vocabulary grounding VLM (3B params, Qwen2.5-3B + MoonViT + Parallel Box Decoder). Image and (frame-sampled) video. 7 operations\\: detect, grounding, point, scene_text, layout, text_grounding, gui_box. Returns fo.Detections or fo.Keypoints (image), {frame\\: label} (video).

Detection,Keypoints,Grounding,Ocr,Layout,Gui,PyTorch,Zero-shot,Image,Video,Vlm,Open-vocabulary,Plugin

MedGemma is a collection of Gemma 3 variants that are trained for performance on medical text and image comprehension

Vlm,Zero-shot,Plugin

C-RADIOv4-H (631M params) - Visual feature extraction model using multi-teacher distillation from SigLIP2, DINOv3, and SAM3. Generates image embeddings and spatial attention features.

Embeddings,Heatmap,Plugin

C-RADIOv4-SO400M (412M params) - Efficient visual feature extraction model. Competitive with ViT-H at lower computational cost.

Embeddings,Heatmap,Plugin

ColQwen is a model based on a novel model architecture and training strategy based on Vision Language Models (VLMs) to efficiently index documents from their visual features. It is a Qwen2.5-VL-3B extension that generates ColBERT- style multi-vector representations of text and images.

Classification,Logits,Embeddings,PyTorch,Visual-document-retrieval,Zero-shot,Plugin

olmOCR-2 is an advanced OCR model from AllenAI that uses Qwen2.5-VL architecture for document text extraction. Returns markdown output with YAML front matter containing document metadata (language, rotation, tables, diagrams). Converts equations to LaTeX and tables to HTML.

Ocr,Document-understanding,Vlm,Vision-language,Plugin

Qwen3-VL-Embedding Specifically designed for multimodal information retrieval and cross-modal understanding, this suite accepts diverse inputs including text, images, screenshots, and videos, as well as inputs containing a mixture of these modalities

Classification,PyTorch,Zero-shot,Video,Embeddings,Plugin

Qwen3-VL-Embedding Specifically designed for multimodal information retrieval and cross-modal understanding, this suite accepts diverse inputs including text, images, screenshots, and videos, as well as inputs containing a mixture of these modalities

Classification,PyTorch,Zero-shot,Video,Embeddings,Plugin

Qwen3-VL is a multimodal vision-language model that processes and understands both text and visual input, enabling it to analyze images, video, and perform advanced reasoning and tasks involving both modalities.

Classification,Detections,PyTorch,Temporal-detections,Zero-shot,Video,Embeddings,Plugin

Qwen3-VL is a multimodal vision-language model that processes and understands both text and visual input, enabling it to analyze images, video, and perform advanced reasoning and tasks involving both modalities.

Classification,Detections,PyTorch,Temporal-detections,Zero-shot,Video,Embeddings,Plugin

Qwen3-VL is a multimodal vision-language model that processes and understands both text and visual input, enabling it to analyze images, video, and perform advanced reasoning and tasks involving both modalities.

Classification,Detections,PyTorch,Temporal-detections,Zero-shot,Video,Embeddings,Plugin

nomic-embed-multimodal-7b is a dense state-of-the-art multimodal embedding model that excels at visual document retrieval tasks.

Classification,Logits,Embeddings,PyTorch,Visual-document-retrieval,Zero-shot,Plugin

nomic-embed-multimodal-3b is a dense state-of-the-art multimodal embedding model that excels at visual document retrieval tasks.

Classification,Logits,Embeddings,PyTorch,Visual-document-retrieval,Zero-shot,Plugin

Online Video Depth Anything (cache_size=16). Estimates temporally-consistent monocular depth for videos using a DINOv2 backbone with a rolling temporal cache. Outputs per-frame fo.Heatmap depth maps normalised to [0, 1].

Depth-estimation,Video,Temporal,Heatmap,Plugin

Online Video Depth Anything (cache_size=8). Lighter variant with a smaller temporal cache; faster and lower memory than c16 at some cost to temporal consistency.

Depth-estimation,Video,Temporal,Heatmap,Plugin

Gemini Vision remote model for VQA via Google Gemini API

Gemini,Vision,Vqa,Multimodal,Plugin

jina-embeddings-v4 is a universal embedding model for multimodal and multilingual retrieval. The model is specially designed for complex document retrieval, including visually rich documents with charts, tables, and illustrations.

Embeddings,PyTorch,Visual-document-retrieval,Zero-shot,Plugin

Moondream 3 (Preview) is an vision language model with a mixture-of-experts architecture (9B total parameters, 2B active).

Detection,Segmentation,Ocr,Vlm,Zero-shot,Plugin

C-RADIOv3-g model (ViT-H/14)

Embeddings,Heatmap,Plugin

C-RADIOv3-H model (ViT-H/16)

Embeddings,Heatmap,Plugin

C-RADIOv3-L model (ViT-L/16))

Embeddings,Heatmap,Plugin

C-RADIOv3-B model (ViT-B/16)

Embeddings,Heatmap,Plugin

Qwen3.5 is a multimodal vision-language model supporting images and videos. Image operations\\: detect, point, classify, vqa, detect_3d. Video operations\\: description, temporal_localization, tracking, ocr, comprehensive, custom.

Detection,Classification,Keypoints,Vqa,Temporal-detections,PyTorch,Zero-shot,Image,Video,3d-detection,Plugin

Qwen3.5 is a multimodal vision-language model supporting images and videos. Image operations\\: detect, point, classify, vqa, detect_3d. Video operations\\: description, temporal_localization, tracking, ocr, comprehensive, custom.

Detection,Classification,Keypoints,Vqa,Temporal-detections,PyTorch,Zero-shot,Image,Video,3d-detection,Plugin

Qwen3.5 is a multimodal vision-language model supporting images and videos. Image operations\\: detect, point, classify, vqa, detect_3d. Video operations\\: description, temporal_localization, tracking, ocr, comprehensive, custom.

Detection,Classification,Keypoints,Vqa,Temporal-detections,PyTorch,Zero-shot,Image,Video,3d-detection,Plugin

Qwen3.5 is a multimodal vision-language model supporting images and videos. Image operations\\: detect, point, classify, vqa, detect_3d. Video operations\\: description, temporal_localization, tracking, ocr, comprehensive, custom.

Detection,Classification,Keypoints,Vqa,Temporal-detections,PyTorch,Zero-shot,Image,Video,3d-detection,Plugin

Qwen3.5 is a multimodal vision-language model supporting images and videos. Image operations\\: detect, point, classify, vqa, detect_3d. Video operations\\: description, temporal_localization, tracking, ocr, comprehensive, custom.

Detection,Classification,Keypoints,Vqa,Temporal-detections,PyTorch,Zero-shot,Image,Video,3d-detection,Plugin

Qwen3.5 is a multimodal vision-language model supporting images and videos. Image operations\\: detect, point, classify, vqa, detect_3d. Video operations\\: description, temporal_localization, tracking, ocr, comprehensive, custom.

Detection,Classification,Keypoints,Vqa,Temporal-detections,PyTorch,Zero-shot,Image,Video,3d-detection,Plugin

Qwen3.5 is a multimodal vision-language model supporting images and videos. Image operations\\: detect, point, classify, vqa, detect_3d. Video operations\\: description, temporal_localization, tracking, ocr, comprehensive, custom.

Detection,Classification,Keypoints,Vqa,Temporal-detections,PyTorch,Zero-shot,Image,Video,3d-detection,Plugin

Gemma 4 E2B is a 2.3B effective parameter multimodal model supporting text, image, video, and audio. Image operations\\: detect, point, classify, vqa. Video operations\\: description, temporal_localization, tracking, ocr, comprehensive, custom.

Detection,Classification,Keypoints,Vqa,Temporal-detections,PyTorch,Zero-shot,Image,Video,Vlm,Plugin

Gemma 4 E4B is a 4.5B effective parameter multimodal model supporting text, image, video, and audio. Image operations\\: detect, point, classify, vqa. Video operations\\: description, temporal_localization, tracking, ocr, comprehensive, custom.

Detection,Classification,Keypoints,Vqa,Temporal-detections,PyTorch,Zero-shot,Image,Video,Vlm,Plugin

Gemma 4 26B-A4B is a 3.8B active parameter (25.2B total) MoE multimodal model supporting text and image. Image operations\\: detect, point, classify, vqa. No video support.

Detection,Classification,Keypoints,Vqa,PyTorch,Zero-shot,Image,Vlm,Plugin

Gemma 4 31B is a 30.7B dense multimodal model supporting text and image. Image operations\\: detect, point, classify, vqa. No video support.

Detection,Classification,Keypoints,Vqa,PyTorch,Zero-shot,Image,Vlm,Plugin

Isaac 0.2 2B (Preview) is an open-source, 2B-parameter model built for real-world applications. Isaac 0.2 is part of Perceptron AI's family of models built to be the intelligence layer for the physical world.

Detection,Ocr,Vlm,Classification,Zero-shot,Plugin

Isaac 0.2 1B is an open-source, 1B-parameter model built for real-world applications. Isaac 0.2 is part of Perceptron AI's family of models built to be the intelligence layer for the physical world.

Detection,Ocr,Vlm,Classification,Zero-shot,Plugin

MolmoPoint-8B is a vision-language model that locates and tracks objects in images and videos by pointing. For images, it returns normalized keypoint coordinates per instance. For videos, it supports frame-level tracking (follows objects over time) and sparse pointing (identifies objects across sampled frames).

Keypoints,Pointing,Tracking,Video,Vlm,Zero-shot,Image-text-to-text,Plugin

MolmoPoint-Vid-4B is a 4B vision-language model specialised for video pointing and tracking. Given a text prompt, it returns keypoint coordinates for each matching object across video frames. Supports frame-level tracking and sparse pointing modes.

Keypoints,Pointing,Tracking,Video,Vlm,Zero-shot,Image-text-to-text,Plugin

MolmoPoint-Img-8B is a vision-language model specialised for pointing at UI elements and objects in screenshots. Given a text prompt, it returns normalized keypoint coordinates for each matching element found.

Keypoints,Pointing,Vlm,Zero-shot,Image-text-to-text,Ui,Screenshots,Plugin

LightOnOCR is a 2.1B-parameter vision-language model optimized for optical character recognition. It uses a chat-based interface to extract text from images with high accuracy across various document types, handwritten text, and scene text.

Ocr,Text-extraction,Vlm,Document-understanding,Plugin

Kimi-VL is an efficient open-source Mixture-of-Experts (MoE) vision-language model (VLM) that offers advanced multimodal reasoning, long-context understanding, and strong agent capabilities—all while activating only 2.8B parameters in its language decoder

Agentic,Detection,Ocr,Vlm,Classification,Zero-shot,Visual-agent,Plugin

Kimi-VL is an efficient open-source Mixture-of-Experts (MoE) vision-language model (VLM) that offers advanced multimodal reasoning, long-context understanding, and strong agent capabilities—all while activating only 2.8B parameters in its language decoder

Agentic,Detection,Segmentation,Ocr,Vlm,Zero-shot,Visual-agent,Plugin

Kimi-VL is an efficient open-source Mixture-of-Experts (MoE) vision-language model (VLM) that offers advanced multimodal reasoning, long-context understanding, and strong agent capabilities—all while activating only 2.8B parameters in its language decoder

Agentic,Detection,Segmentation,Ocr,Vlm,Zero-shot,Visual-agent,Plugin

SigLIP 2 extends the pretraining objective of SigLIP with prior, independently developed techniques into a unified recipe, for improved semantic understanding, localization, and dense features.

Classification,Logits,Embeddings,PyTorch,Clip,Zero-shot,Plugin

SigLIP 2 extends the pretraining objective of SigLIP with prior, independently developed techniques into a unified recipe, for improved semantic understanding, localization, and dense features.

Classification,Logits,Embeddings,PyTorch,Clip,Zero-shot,Plugin

SigLIP 2 extends the pretraining objective of SigLIP with prior, independently developed techniques into a unified recipe, for improved semantic understanding, localization, and dense features.

Classification,Logits,Embeddings,PyTorch,Clip,Zero-shot,Plugin

SigLIP 2 extends the pretraining objective of SigLIP with prior, independently developed techniques into a unified recipe, for improved semantic understanding, localization, and dense features.

Classification,Logits,Embeddings,PyTorch,Clip,Zero-shot,Plugin

SigLIP 2 extends the pretraining objective of SigLIP with prior, independently developed techniques into a unified recipe, for improved semantic understanding, localization, and dense features.

Classification,Logits,Embeddings,PyTorch,Clip,Zero-shot,Plugin

SigLIP 2 extends the pretraining objective of SigLIP with prior, independently developed techniques into a unified recipe, for improved semantic understanding, localization, and dense features.

Classification,Logits,Embeddings,PyTorch,Clip,Zero-shot,Plugin

SigLIP 2 extends the pretraining objective of SigLIP with prior, independently developed techniques into a unified recipe, for improved semantic understanding, localization, and dense features.

Classification,Logits,Embeddings,PyTorch,Clip,Zero-shot,Plugin

SigLIP 2 extends the pretraining objective of SigLIP with prior, independently developed techniques into a unified recipe, for improved semantic understanding, localization, and dense features.

Classification,Logits,Embeddings,PyTorch,Clip,Zero-shot,Plugin

SigLIP 2 extends the pretraining objective of SigLIP with prior, independently developed techniques into a unified recipe, for improved semantic understanding, localization, and dense features.

Classification,Logits,Embeddings,PyTorch,Clip,Zero-shot,Plugin

SigLIP 2 extends the pretraining objective of SigLIP with prior, independently developed techniques into a unified recipe, for improved semantic understanding, localization, and dense features.

Classification,Logits,Embeddings,PyTorch,Clip,Zero-shot,Plugin

SigLIP 2 extends the pretraining objective of SigLIP with prior, independently developed techniques into a unified recipe, for improved semantic understanding, localization, and dense features.

Classification,Logits,Embeddings,PyTorch,Clip,Zero-shot,Plugin

SigLIP 2 extends the pretraining objective of SigLIP with prior, independently developed techniques into a unified recipe, for improved semantic understanding, localization, and dense features.

Classification,Logits,Embeddings,PyTorch,Clip,Zero-shot,Plugin

SigLIP 2 extends the pretraining objective of SigLIP with prior, independently developed techniques into a unified recipe, for improved semantic understanding, localization, and dense features.

Classification,Logits,Embeddings,PyTorch,Clip,Zero-shot,Plugin

SigLIP 2 extends the pretraining objective of SigLIP with prior, independently developed techniques into a unified recipe, for improved semantic understanding, localization, and dense features.

Classification,Logits,Embeddings,PyTorch,Clip,Zero-shot,Plugin

SigLIP 2 extends the pretraining objective of SigLIP with prior, independently developed techniques into a unified recipe, for improved semantic understanding, localization, and dense features.

Classification,Logits,Embeddings,PyTorch,Clip,Zero-shot,Plugin

SigLIP 2 extends the pretraining objective of SigLIP with prior, independently developed techniques into a unified recipe, for improved semantic understanding, localization, and dense features.

Classification,Logits,Embeddings,PyTorch,Clip,Zero-shot,Plugin

SigLIP 2 extends the pretraining objective of SigLIP with prior, independently developed techniques into a unified recipe, for improved semantic understanding, localization, and dense features.

Classification,Logits,Embeddings,PyTorch,Clip,Zero-shot,Plugin

Molmo2 is a family of open vision-language models developed by the Allen Institute for AI (Ai2) that support image, video and multi-image understanding and grounding.

PyTorch,Keypoints,Zero-shot,Video,Plugin

Molmo2 is a family of open vision-language models developed by the Allen Institute for AI (Ai2) that support image, video and multi-image understanding and grounding.

PyTorch,Keypoints,Zero-shot,Video,Plugin

Molmo2 is a family of open vision-language models developed by the Allen Institute for AI (Ai2) that support image, video and multi-image understanding and grounding.

PyTorch,Keypoints,Zero-shot,Video,Plugin

Molmo2 is a family of open vision-language models developed by the Allen Institute for AI (Ai2) that support image, video and multi-image understanding and grounding.

PyTorch,Keypoints,Zero-shot,Video,Plugin

SAM3 (Segment Anything Model 3) performs promptable segmentation on images using text or visual prompts. Supports concept segmentation (find all instances), visual segmentation (specific instances), automatic segmentation, and visual embeddings.

Embeddings,PyTorch,Zero-shot,Segmentation,Segment-anything,Detections,Instance-segmentation,Visual-prompting,Plugin

FastVLM is a vision-language model from Apple that excels at visual question answering and image classification tasks. It supports both zero-shot classification and open-ended VQA with customizable prompts.

Vision-language,Vqa,Multimodal,Transformers,PyTorch,Plugin

FastVLM is a vision-language model from Apple that excels at visual question answering and image classification tasks. It supports both zero-shot classification and open-ended VQA with customizable prompts.

Vision-language,Vqa,Multimodal,Transformers,PyTorch,Plugin

FastVLM is a vision-language model from Apple that excels at visual question answering and image classification tasks. It supports both zero-shot classification and open-ended VQA with customizable prompts.

Vision-language,Vqa,Multimodal,Transformers,PyTorch,Plugin

GLM-OCR is a vision-language model for document understanding. Supports text recognition, formula recognition, table recognition, and custom structured extraction via JSON prompts.

Ocr,Document-understanding,Visual-document-retrieval,Plugin

GUI-Actor is Coordinate-Free Visual Grounding for GUI Agents

Keypoints,Visual-agent,Plugin

GUI-Actor is Coordinate-Free Visual Grounding for GUI Agents

Keypoints,Visual-agent,Plugin

MedGemma 1.5 4B is an updated version of the MedGemma 1 4B model. MedGemma is a collection of Gemma 3 variants that are trained for performance on medical text and image comprehension

Vlm,Zero-shot,Plugin

Isaac 0.1 is an open-source, 2B-parameter model built for real-world applications. Isaac 0.1 is the first in Perceptron AI's family of models built to be the intelligence layer for the physical world.

Detection,Ocr,Vlm,Classification,Zero-shot,Plugin

SHARP is an approach to photorealistic view synthesis from a single image. Given a single photograph, SHARP regresses the parameters of a 3D Gaussian representation of the depicted scene.

Threed,Plugin

MinerU2.5 is a 1.2B-parameter vision-language model for document parsing. It adopts a two-stage parsing strategy\\: first conducting efficient global layout analysis on downsampled images, then performing fine-grained content recognition on native-resolution crops for text, formulas, and tables.

Detection,Ocr,Vlm,Plugin

Llama Nemotron Nano VL is a leading document intelligence vision language model (VLMs) that enables the ability to query and summarize images from the physical or virtual world.

Detection,Ocr,Vlm,Classification,Zero-shot,Visual-agent,Plugin

Visual Geometry Grounded Transformer (VGGT) is a feed-forward neural network that directly infers all key 3D attributes of a scene.

Depth,Threed,Keypoints,Plugin

MedSigLIP is a variant of SigLIP that is trained to encode medical images and text into a common embedding space.

Classification,Logits,Embeddings,PyTorch,Clip,Zero-shot,Plugin

Nanonets-OCR2 are image-to-markdown OCR models that go far beyond traditional text extraction. It transforms documents into structured markdown with intelligent content recognition and semantic tagging, making it ideal for downstream processing by Large Language Models (LLMs).

Detection,Ocr,Vlm,Plugin

DeepSeek-OCR is an open-source vision-language model (VLM) developed by DeepSeek to perform optical character recognition (OCR) and context compression for long and complex documents

Detection,Ocr,Vlm,Plugin

Kosmos-2.5 is a multimodal literate model for machine reading of text-intensive images.

Detection,Ocr,Vlm,Plugin

The ModernVBERT suite is a suite of compact 250M-parameter vision-language encoders. BiModernVBERT is the bi-encoder version that is fine-tuned for visual document retrieval tasks.

Classification,Logits,Embeddings,PyTorch,Visual-document-retrieval,Zero-shot,Plugin

The ModernVBERT suite is a suite of compact 250M-parameter vision-language encoders. ColModernVBERT is the late-interaction version that is fine-tuned for visual document retrieval tasks, the most performant model on this task.

Classification,Logits,Embeddings,PyTorch,Visual-document-retrieval,Zero-shot,Plugin

UI-TARS-1.5 is an open-source multimodal agent capable of effectively performing diverse tasks within virtual worlds.

Detection,Ocr,Vlm,Classification,Zero-shot,Visual-agent,Plugin

ColPali is a model based on a novel model architecture and training strategy based on Vision Language Models (VLMs) to efficiently index documents from their visual features. It is a PaliGemma-3B extension that generates ColBERT- style multi-vector representations of text and images

Classification,Logits,Embeddings,PyTorch,Visual-document-retrieval,Zero-shot,Plugin

PaliGemma 2 mix checkpoints are fine-tuned on a diverse set of tasks and are ready to use out of the box. These tasks include short and long captioning, optical character recognition, question answering, object detection and segmentation, and more.

Detection,Segmentation,Ocr,Vlm,Zero-shot,Plugin

PaliGemma 2 mix checkpoints are fine-tuned on a diverse set of tasks and are ready to use out of the box. These tasks include short and long captioning, optical character recognition, question answering, object detection and segmentation, and more.

Detection,Segmentation,Ocr,Vlm,Zero-shot,Plugin

PaliGemma 2 mix checkpoints are fine-tuned on a diverse set of tasks and are ready to use out of the box. These tasks include short and long captioning, optical character recognition, question answering, object detection and segmentation, and more.

Detection,Segmentation,Ocr,Vlm,Zero-shot,Plugin

PaliGemma 2 mix checkpoints are fine-tuned on a diverse set of tasks and are ready to use out of the box. These tasks include short and long captioning, optical character recognition, question answering, object detection and segmentation, and more.

Detection,Segmentation,Ocr,Vlm,Zero-shot,Plugin

PaliGemma 2 mix checkpoints are fine-tuned on a diverse set of tasks and are ready to use out of the box. These tasks include short and long captioning, optical character recognition, question answering, object detection and segmentation, and more.

Detection,Segmentation,Ocr,Vlm,Zero-shot,Plugin

PaliGemma 2 mix checkpoints are fine-tuned on a diverse set of tasks and are ready to use out of the box. These tasks include short and long captioning, optical character recognition, question answering, object detection and segmentation, and more.

Detection,Segmentation,Ocr,Vlm,Zero-shot,Plugin

MiniCPM-V 4.5 is the latest and most capable model in the MiniCPM-V series. The model is built on Qwen3-8B and SigLIP2-400M with a total of 8B parameters.

Detection,Keypoints,Ocr,Vlm,Classification,Zero-shot,Plugin

Moondream is a small vision language model designed to run efficiently on edge devices.

Detection,Segmentation,Ocr,Vlm,Zero-shot,Plugin

Florence-2 is a vision foundation model with a unified, prompt-based representation for a variety of computer vision and vision-language tasks (https\\://arxiv.org/abs/2311.06242).

Detection,Segmentation,Ocr,Vlm,Zero-shot,Plugin

Florence-2 is a vision foundation model with a unified, prompt-based representation for a variety of computer vision and vision-language tasks (https\\://arxiv.org/abs/2311.06242).

Detection,Segmentation,Ocr,Vlm,Zero-shot,Plugin

Florence-2 is a vision foundation model with a unified, prompt-based representation for a variety of computer vision and vision-language tasks (https\\://arxiv.org/abs/2311.06242).

Detection,Segmentation,Ocr,Vlm,Zero-shot,Plugin

Florence-2 is a vision foundation model with a unified, prompt-based representation for a variety of computer vision and vision-language tasks (https\\://arxiv.org/abs/2311.06242).

Detection,Segmentation,Ocr,Vlm,Zero-shot,Plugin

ShowUI is a lightweight (2B) vision-language-action model designed for GUI agents.

Detection,Ocr,Vlm,Classification,Zero-shot,Visual-agent,Plugin

MiMo-VL-7B is a compact yet powerful vision-language model developed through extensive pre-training and reinforcement learning to achieve state-of-the-art performance on a variety of visual-language tasks.

Detection,Ocr,Vlm,Classification,Zero-shot,Visual-agent,Plugin

MiMo-VL-7B is a compact yet powerful vision-language model developed through extensive pre-training and reinforcement learning to achieve state-of-the-art performance on a variety of visual-language tasks.

Detection,Ocr,Vlm,Classification,Zero-shot,Visual-agent,Plugin

MiMo-VL-7B is a compact yet powerful vision-language model developed through extensive pre-training and reinforcement learning to achieve state-of-the-art performance on a variety of visual-language tasks.

Detection,Ocr,Vlm,Classification,Zero-shot,Visual-agent,Plugin

OS-Atlas provides a series of models specifically designed for GUI agents.

Detection,Ocr,Vlm,Classification,Zero-shot,Visual-agent,Plugin

Qwen2.5-VL is the multimodal large language model series developed by Qwen team, Alibaba Cloud.

Detection,Ocr,Vlm,Classification,Zero-shot,Plugin

Qwen2.5-VL is the multimodal large language model series developed by Qwen team, Alibaba Cloud.

Detection,Segmentation,Ocr,Vlm,Zero-shot,Plugin

Qwen2.5-VL is the multimodal large language model series developed by Qwen team, Alibaba Cloud.

Detection,Segmentation,Ocr,Vlm,Zero-shot,Plugin

Qwen2.5-VL is the multimodal large language model series developed by Qwen team, Alibaba Cloud.

Detection,Segmentation,Ocr,Vlm,Zero-shot,Plugin

Qwen2.5-VL is the multimodal large language model series developed by Qwen team, Alibaba Cloud.

Detection,Segmentation,Ocr,Vlm,Zero-shot,Plugin

Qwen2.5-VL is the multimodal large language model series developed by Qwen team, Alibaba Cloud.

Detection,Segmentation,Ocr,Vlm,Zero-shot,Plugin

Qwen2.5-VL is the multimodal large language model series developed by Qwen team, Alibaba Cloud.

Detection,Segmentation,Ocr,Vlm,Zero-shot,Plugin

Qwen2.5-VL is the multimodal large language model series developed by Qwen team, Alibaba Cloud.

Detection,Segmentation,Ocr,Vlm,Zero-shot,Plugin

vdr-2b-v1 is an english only embedding model designed for visual document retrieval. It encodes document page screenshots into dense single-vector representations, this will effectively allow to search and query visually rich multilingual documents without the need for any OCR, data extraction pipelines, and chunking.

Embeddings,Ocr,Vlm,Document-retrieval,Plugin

vdr-2b-multi-v1 is a multilingual embedding model designed for visual document retrieval across multiple languages and domains. It encodes document page screenshots into dense single-vector representations, this will effectively allow to search and query visually rich multilingual documents without the need for any OCR, data extraction pipelines, and chunking. It's trained on 🇮🇹 Italian, 🇪🇸 Spanish, 🇬🇧 English, 🇫🇷 French and 🇩🇪 German

Embeddings,Ocr,Vlm,Document-retrieval,Plugin