RADIO L

🚀 AM-RADIO: Reduce All Domains Into One

AM-RADIO is a vision foundation model that aims to reduce all domains into one. It provides a unified approach for various computer - vision tasks.

Authors: Mike Ranzinger, Greg Heinrich, Jan Kautz, Pavlo Molchanov

VIDIA Research

[AM-RADIO Paper] [PHI-S Paper] [BibTex][GitHub examples] [Tech report on v2.5]

🚀 Quick Start

HuggingFace Hub

You can pull the model from a Python script:

import torch
from PIL import Image
from transformers import AutoModel, CLIPImageProcessor

hf_repo = "nvidia/RADIO-L"

image_processor = CLIPImageProcessor.from_pretrained(hf_repo)
model = AutoModel.from_pretrained(hf_repo, trust_remote_code=True)
model.eval().cuda()

image = Image.open('./assets/radio.png').convert('RGB')
pixel_values = image_processor(images=image, return_tensors='pt', do_resize=True).pixel_values
pixel_values = pixel_values.cuda()

summary, features = model(pixel_values)

💻 Usage Examples

Basic Usage

The code above in the HuggingFace Hub section shows the basic way to load the model and get the output.

Advanced Usage

RADIO will return a tuple with two tensors. The summary is similar to the cls_token in ViT and is meant to represent the general concept of the entire image. It has shape $(B,C)$ with $B$ being the batch dimension, and $C$ being some number of channels. The spatial_features represent more localized content which should be suitable for dense tasks such as semantic segmentation, or for integration into an LLM. It has shape $(B,T,D)$ with $T$ being the flattened spatial tokens, and $D$ being the channels for spatial features. Note that $C \neq D$ in general.

Converting to a spatial tensor format can be done using the downsampling size of the model, combined with the input tensor shape. For 'radio_v1', the patch size is 14.

from einops import rearrange
spatial_features = rearrange(spatial_features, 'b (h w) d -> b d h w', h=x.shape[-2] // patch_size, w=x.shape[-1] // patch_size)

The resulting tensor will have shape $(B,D,H,W)$, as is typically seen with computer vision models.

📚 Documentation

RADIOv2.5 Notes

See the RADIOv2.5 technical report.

📄 License

RADIO code and weights are released under the NSCLv1 License.

Citing RADIO

If you find this repository useful, please consider giving a star and citation:

@InProceedings{Ranzinger_2024_CVPR,
    author    = {Ranzinger, Mike and Heinrich, Greg and Kautz, Jan and Molchanov, Pavlo},
    title     = {AM-RADIO: Agglomerative Vision Foundation Model Reduce All Domains Into One},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2024},
    pages     = {12490-12500}
}

@misc{ranzinger2024phisdistributionbalancinglabelfree,
      title={PHI-S: Distribution Balancing for Label-Free Multi-Teacher Distillation}, 
      author={Mike Ranzinger and Jon Barker and Greg Heinrich and Pavlo Molchanov and Bryan Catanzaro and Andrew Tao},
      year={2024},
      eprint={2410.01680},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2410.01680}, 
}