argus-lite-torch#

Argus-Lite (38M) multi-task perception model on a frozen EUPE-ViT-S backbone for COCO detection, ImageNet classification, ADE20K semantic segmentation and indoor metric depth.

Details

  • Model name: argus-lite-torch

  • Model source: https://huggingface.co/phanerozoic/argus-lite

  • Model author: Charles Norton (phanerozoic)

  • Model license: FAIR Noncommercial Research License

  • Model size: 145.28 MB

  • Exposes embeddings? yes

  • Tags: detection, classification, segmentation, depth, embeddings, torch, transformer

Requirements

  • Packages: torch, torchvision>=0.16, transformers

  • CPU support

    • yes

  • GPU support

    • yes

Example usage

 1import fiftyone as fo
 2import fiftyone.zoo as foz
 3
 4dataset = foz.load_zoo_dataset(
 5    "coco-2017",
 6    split="validation",
 7    dataset_name=fo.get_default_dataset_name(),
 8    max_samples=50,
 9    shuffle=True,
10)
11
12model = foz.load_zoo_model("argus-lite-torch")
13
14dataset.apply_model(model, label_field="predictions")
15
16session = fo.launch_app(dataset)