Home

Awesome

Visual Attention Network (VAN) paper pdf

ImageNet-22K pretrained models will be Released before July 20, 2022.

This is a PyTorch implementation of VAN proposed by our paper "Visual Attention Network".

Comparsion

Figure 1: Compare with different vision backbones on ImageNet-1K validation set.

Citation:

@article{guo2022visual,
  title={Visual Attention Network},
  author={Guo, Meng-Hao and Lu, Cheng-Ze and Liu, Zheng-Ning and Cheng, Ming-Ming and Hu, Shi-Min},
  journal={arXiv preprint arXiv:2202.09741},
  year={2022}
}

News:

2022.02.22 Release paper on ArXiv and code on github.

2022.02.25 Supported by Jimm

2022.03.15 Supported by Hugging Face.

2022.04 Supported by PaddleCls.

2022.05 Supported by OpenMMLab.

For More Code, please refer to Paper with code.

2022.07.08 Update paper on ArXiv. (ImageNet-22K results, SOTA for panoptic segmentation (58.2 PQ). Segmentation models are available.

Abstract:

While originally designed for natural language processing (NLP) tasks, the self-attention mechanism has recently taken various computer vision areas by storm. However, the 2D nature of images brings three challenges for applying self-attention in computer vision. (1) Treating images as 1D sequences neglects their 2D structures. (2) The quadratic complexity is too expensive for high-resolution images. (3) It only captures spatial adaptability but ignores channel adaptability. In this paper, we propose a novel large kernel attention (LKA) module to enable self-adaptive and long-range correlations in self-attention while avoiding the above issues. We further introduce a novel neural network based on LKA, namely Visual Attention Network (VAN). While extremely simple and efficient, VAN outperforms the state-of-the-art vision transformers (ViTs) and convolutional neural networks (CNNs) with a large margin in extensive experiments, including image classification, object detection, semantic segmentation, instance segmentation, etc.

Decomposition

Figure 2: Decomposition diagram of large-kernel convolution. A standard convolution can be decomposed into three parts: a depth-wise convolution (DW-Conv), a depth-wise dilation convolution (DW-D-Conv) and a 1×1 convolution (1×1 Conv).

LKA

Figure 3: The structure of different modules: (a) the proposed Large Kernel Attention (LKA); (b) non-attention module; (c) the self-attention module (d) a stage of our Visual Attention Network (VAN). CFF means convolutional feed-forward network. The difference between (a) and (b) is the element-wise multiply. It is worth noting that (c) is designed for 1D sequences. .

Image Classification

Data prepare: ImageNet with the following folder structure.

│imagenet/
├──train/
│  ├── n01440764
│  │   ├── n01440764_10026.JPEG
│  │   ├── n01440764_10027.JPEG
│  │   ├── ......
│  ├── ......
├──val/
│  ├── n01440764
│  │   ├── ILSVRC2012_val_00000293.JPEG
│  │   ├── ILSVRC2012_val_00002138.JPEG
│  │   ├── ......
│  ├── ......

2. VAN Models (IN-1K)

Model#Params(M)GFLOPsTop1 Acc(%)Download
VAN-B04.10.975.4Google Drive, Tsinghua Cloud, Hugging Face 🤗
VAN-B113.92.581.1Google Drive, Tsinghua Cloud, Hugging Face 🤗
VAN-B226.65.082.8Google Drive, Tsinghua Cloud,Hugging Face 🤗,
VAN-B344.89.083.9Google Drive, Tsinghua Cloud, Hugging Face 🤗
VAN-B4TODOTODOTODOTODO

3.Requirement

1. Pytorch >= 1.7
2. timm == 0.4.12

4. Train

We use 8 GPUs for training by default. Run command (It has been writen in train.sh):

MODEL=van_tiny # van_{tiny, small, base, large}
DROP_PATH=0.1 # drop path rates [0.1, 0.1, 0.1, 0.2] for [tiny, small, base, large]
CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 bash distributed_train.sh 8 /path/to/imagenet \
	  --model $MODEL -b 128 --lr 1e-3 --drop-path $DROP_PATH

5. Validate

Run command (It has been writen in eval.sh) as:

MODEL=van_tiny # van_{tiny, small, base, large}
python3 validate.py /path/to/imagenet --model $MODEL \
  --checkpoint /path/to/model -b 128

6.Acknowledgment

Our implementation is mainly based on pytorch-image-models and PoolFormer. Thanks for their authors.

LICENSE

This repo is under the Apache-2.0 license. For commercial use, please contact the authors.