c-radio-v4-h-torch#

C-RADIOv4-H (631M params) vision foundation model combining SigLIP2, DINOv3, and SAM3 for embeddings and spatial attention.

Details

  • Model name: c-radio-v4-h-torch

  • Model source: NVlabs/RADIO

  • Model author: NVIDIA

  • Model license: NVIDIA Open Model License Agreement

  • Model size: 2.35 GB

  • Exposes embeddings? yes

  • Tags: embeddings, heatmap, torch, transformer, nvidia, foundation

Requirements

  • Packages: torch, torchvision, transformers, timm, einops, open-clip-torch, scikit-learn, scipy

  • 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("c-radio-v4-h-torch")
13
14dataset.apply_model(model, label_field="predictions")
15
16session = fo.launch_app(dataset)