Home

Awesome

EfficientViT: Multi-Scale Linear Attention for High-Resolution Dense Prediction [paper]

Efficient vision foundation models for high-resolution generation and perception.

Content

Deep Compression Autoencoder for Efficient High-Resolution Diffusion Models [paper] [readme]

Deep Compression Autoencoder (DC-AE) is a new family of high-spatial compression autoencoders with a spatial compression ratio of up to 128 while maintaining reconstruction quality. It accelerates all latent diffusion models regardless of the diffusion model architecture.

Updates

<p align="left"> <img src="assets/uvit_2b_imagenet_512px.png" width="1200"> </p> <p align="left"> <img src="assets/diffusion_scaling_up.jpg" width="300"> </p>

Demo

demo

<p align="center"> <b> Figure 1: We address the reconstruction accuracy drop of high spatial-compression autoencoders. </p>

demo

<p align="center"> <b> Figure 2: DC-AE speeds up latent diffusion models. </p> <p align="left"> <img src="https://huggingface.co/mit-han-lab/dc-ae-f64c128-in-1.0/resolve/main/assets/Sana-0.6B-laptop.png" width="1200"> </p> <p align="center"> <img src="https://huggingface.co/mit-han-lab/dc-ae-f64c128-in-1.0/resolve/main/assets/dc_ae_sana.jpg" width="1200"> </p> <p align="center"> <b> Figure 3: DC-AE enables efficient text-to-image generation on the laptop. For more details, please check our text-to-image diffusion model <a href="https://nvlabs.github.io/Sana/">SANA</a>. </p>

EfficientViT-SAM: Accelerated Segment Anything Model Without Accuracy Loss [paper] [online demo] [readme]

EfficientViT-SAM is a new family of accelerated segment anything models by replacing SAM's heavy image encoder with EfficientViT. It delivers a 48.9x measured TensorRT speedup on A100 GPU over SAM-ViT-H without sacrificing accuracy.

<p align="left"> <img src="https://huggingface.co/mit-han-lab/efficientvit-sam/resolve/main/sam_zero_shot_coco_mAP.png" width="500"> </p>

EfficientViT-Classification [paper] [readme]

Efficient image classification models with EfficientViT backbones.

<p align="left"> <img src="https://huggingface.co/han-cai/efficientvit-cls/resolve/main/efficientvit_cls_results.png" width="600"> </p>

EfficientViT-Segmentation [paper] [readme]

Efficient semantic segmantation models with EfficientViT backbones.

demo

EfficientViT-GazeSAM [readme]

Gaze-prompted image segmentation models capable of running in real time with TensorRT on an NVIDIA RTX 4070.

GazeSAM demo

News

If you are interested in getting updates, please join our mailing list here.

Getting Started

conda create -n efficientvit python=3.10
conda activate efficientvit
pip install -U -r requirements.txt

Third-Party Implementation/Integration

Contact

Han Cai

Reference

If EfficientViT or EfficientViT-SAM or DC-AE is useful or relevant to your research, please kindly recognize our contributions by citing our paper:

@inproceedings{cai2023efficientvit,
  title={Efficientvit: Lightweight multi-scale attention for high-resolution dense prediction},
  author={Cai, Han and Li, Junyan and Hu, Muyan and Gan, Chuang and Han, Song},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  pages={17302--17313},
  year={2023}
}
@article{zhang2024efficientvit,
  title={EfficientViT-SAM: Accelerated Segment Anything Model Without Performance Loss},
  author={Zhang, Zhuoyang and Cai, Han and Han, Song},
  journal={arXiv preprint arXiv:2402.05008},
  year={2024}
}
@article{chen2024deep,
  title={Deep Compression Autoencoder for Efficient High-Resolution Diffusion Models},
  author={Chen, Junyu and Cai, Han and Chen, Junsong and Xie, Enze and Yang, Shang and Tang, Haotian and Li, Muyang and Lu, Yao and Han, Song},
  journal={arXiv preprint arXiv:2410.10733},
  year={2024}
}