Home

Awesome

RepViT-SAM: Towards Real-Time Segmenting Anything

RepViT: Revisiting Mobile CNN From ViT Perspective

Official PyTorch implementation of RepViT-SAM and RepViT. CVPR 2024.

<p align="center"> <img src="sam/figures/comparison.png" width=80%> <br> Models are deployed on iPhone 12 with Core ML Tools to get latency. </p> <p align="center"> <img src="figures/latency.png" width=70%> <br> Models are trained on ImageNet-1K and deployed on iPhone 12 with Core ML Tools to get latency. </p>

RepViT-SAM: Towards Real-Time Segmenting Anything.
Ao Wang, Hui Chen, Zijia Lin, Jungong Han, and Guiguang Ding
[arXiv] [Project Page]

<details> <summary> <font size="+1">Abstract</font> </summary> Segment Anything Model (SAM) has shown impressive zero-shot transfer performance for various computer vision tasks recently. However, its heavy computation costs remain daunting for practical applications. MobileSAM proposes to replace the heavyweight image encoder in SAM with TinyViT by employing distillation, which results in a significant reduction in computational requirements. However, its deployment on resource-constrained mobile devices still encounters challenges due to the substantial memory and computational overhead caused by self-attention mechanisms. Recently, RepViT achieves the state-of-the-art performance and latency trade-off on mobile devices by incorporating efficient architectural designs of ViTs into CNNs. Here, to achieve real-time segmenting anything on mobile devices, following, we replace the heavyweight image encoder in SAM with RepViT model, ending up with the RepViT-SAM model. Extensive experiments show that RepViT-SAM can enjoy significantly better zero-shot transfer capability than MobileSAM, along with nearly $10\times$ faster inference speed. </details> <br/>

RepViT: Revisiting Mobile CNN From ViT Perspective.
Ao Wang, Hui Chen, Zijia Lin, Jungong Han, and Guiguang Ding
[arXiv]

<details> <summary> <font size="+1">Abstract</font> </summary> Recently, lightweight Vision Transformers (ViTs) demonstrate superior performance and lower latency compared with lightweight Convolutional Neural Networks (CNNs) on resource-constrained mobile devices. This improvement is usually attributed to the multi-head self-attention module, which enables the model to learn global representations. However, the architectural disparities between lightweight ViTs and lightweight CNNs have not been adequately examined. In this study, we revisit the efficient design of lightweight CNNs and emphasize their potential for mobile devices. We incrementally enhance the mobile-friendliness of a standard lightweight CNN, specifically MobileNetV3, by integrating the efficient architectural choices of lightweight ViTs. This ends up with a new family of pure lightweight CNNs, namely RepViT. Extensive experiments show that RepViT outperforms existing state-of-the-art lightweight ViTs and exhibits favorable latency in various vision tasks. On ImageNet, RepViT achieves over 80\% top-1 accuracy with 1ms latency on an iPhone 12, which is the first time for a lightweight model, to the best of our knowledge. Our largest model, RepViT-M2.3, obtains 83.7\% accuracy with only 2.3ms latency. </details> <br/> <br/>

UPDATES 🔥

<br/>

Classification on ImageNet-1K

Models

ModelTop-1 (300 / 450)#paramsMACsLatencyCkptCore MLLog
M0.978.7 / 79.15.1M0.8G0.9ms300e / 450e300e / 450e300e / 450e
M1.080.0 / 80.36.8M1.1G1.0ms300e / 450e300e / 450e300e / 450e
M1.180.7 / 81.28.2M1.3G1.1ms300e / 450e300e / 450e300e / 450e
M1.582.3 / 82.514.0M2.3G1.5ms300e / 450e300e / 450e300e / 450e
M2.383.3 / 83.722.9M4.5G2.3ms300e / 450e300e / 450e300e / 450e

Tips: Convert a training-time RepViT into the inference-time structure

from timm.models import create_model
import utils

model = create_model('repvit_m0_9')
utils.replace_batchnorm(model)

Latency Measurement

The latency reported in RepViT for iPhone 12 (iOS 16) uses the benchmark tool from XCode 14. For example, here is a latency measurement of RepViT-M0.9:

Tips: export the model to Core ML model

python export_coreml.py --model repvit_m0_9 --ckpt pretrain/repvit_m0_9_distill_300e.pth

Tips: measure the throughput on GPU

python speed_gpu.py --model repvit_m0_9

ImageNet

Prerequisites

conda virtual environment is recommended.

conda create -n repvit python=3.8
pip install -r requirements.txt

Data preparation

Download and extract ImageNet train and val images from http://image-net.org/. The training and validation data are expected to be in the train folder and val folder respectively:

|-- /path/to/imagenet/
    |-- train
    |-- val

Training

To train RepViT-M0.9 on an 8-GPU machine:

python -m torch.distributed.launch --nproc_per_node=8 --master_port 12346 --use_env main.py --model repvit_m0_9 --data-path ~/imagenet --dist-eval

Tips: specify your data path and model name!

Testing

For example, to test RepViT-M0.9:

python main.py --eval --model repvit_m0_9 --resume pretrain/repvit_m0_9_distill_300e.pth --data-path ~/imagenet

Downstream Tasks

Object Detection and Instance Segmentation<br> Semantic Segmentation

Acknowledgement

Classification (ImageNet) code base is partly built with LeViT, PoolFormer and EfficientFormer.

The detection and segmentation pipeline is from MMCV (MMDetection and MMSegmentation).

Thanks for the great implementations!

Citation

If our code or models help your work, please cite our paper:

@inproceedings{wang2024repvit,
  title={Repvit: Revisiting mobile cnn from vit perspective},
  author={Wang, Ao and Chen, Hui and Lin, Zijia and Han, Jungong and Ding, Guiguang},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={15909--15920},
  year={2024}
}

@misc{wang2023repvitsam,
      title={RepViT-SAM: Towards Real-Time Segmenting Anything}, 
      author={Ao Wang and Hui Chen and Zijia Lin and Jungong Han and Guiguang Ding},
      year={2023},
      eprint={2312.05760},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}