Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

https://github.com/jeonsworld/ViT-pytorch

Pytorch reimplementation of the Vision Transformer (An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale)
https://github.com/jeonsworld/ViT-pytorch

Last synced: about 2 months ago
JSON representation

Pytorch reimplementation of the Vision Transformer (An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale)

Host: GitHub
URL: https://github.com/jeonsworld/ViT-pytorch
Owner: jeonsworld
License: mit
Created: 2020-11-03T09:50:50.000Z (over 3 years ago)
Default Branch: main
Last Pushed: 2022-06-07T07:14:33.000Z (about 2 years ago)
Last Synced: 2024-02-04T23:11:46.610Z (5 months ago)
Language: Jupyter Notebook
Homepage:
Size: 20.8 MB
Stars: 1,721
Watchers: 13
Forks: 349
Open Issues: 28
Metadata Files:
- Readme: README.md
- License: LICENSE

Lists

awesome-image-classification - unofficial-pytorch: https://github.com/jeonsworld/ViT-pytorch
my-awesome-stars - jeonsworld/ViT-pytorch - Pytorch reimplementation of the Vision Transformer (An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale) (Jupyter Notebook)
Visual-Transformer-Paper-Summary - ViT-pytorch

README

        # Vision Transformer

Pytorch reimplementation of [Google's repository for the ViT model](https://github.com/google-research/vision_transformer) that was released with the paper [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929) by Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby.

This paper show that Transformers applied directly to image patches and pre-trained on large datasets work really well on image recognition task.

![fig1](./img/figure1.png)

Vision Transformer achieve State-of-the-Art in image recognition task with standard Transformer encoder and fixed-size patches. In order to perform classification, author use the standard approach of adding an extra learnable "classification token" to the sequence.

![fig2](./img/figure2.png)

## Usage

### 1. Download Pre-trained model (Google's Official Checkpoint)

* [Available models](https://console.cloud.google.com/storage/vit_models/): ViT-B_16(**85.8M**), R50+ViT-B_16(**97.96M**), ViT-B_32(**87.5M**), ViT-L_16(**303.4M**), ViT-L_32(**305.5M**), ViT-H_14(**630.8M**)

  * imagenet21k pre-train models

    * ViT-B_16, ViT-B_32, ViT-L_16, ViT-L_32, ViT-H_14

  * imagenet21k pre-train + imagenet2012 fine-tuned models

    * ViT-B_16-224, ViT-B_16, ViT-B_32, ViT-L_16-224, ViT-L_16, ViT-L_32

  * Hybrid Model([Resnet50](https://github.com/google-research/big_transfer) + Transformer)

    * R50-ViT-B_16

```

# imagenet21k pre-train

wget https://storage.googleapis.com/vit_models/imagenet21k/{MODEL_NAME}.npz

# imagenet21k pre-train + imagenet2012 fine-tuning

wget https://storage.googleapis.com/vit_models/imagenet21k+imagenet2012/{MODEL_NAME}.npz

```

### 2. Train Model

```

python3 train.py --name cifar10-100_500 --dataset cifar10 --model_type ViT-B_16 --pretrained_dir checkpoint/ViT-B_16.npz

```

CIFAR-10 and CIFAR-100 are automatically download and train. In order to use a different dataset you need to customize [data_utils.py](./utils/data_utils.py).

The default batch size is 512. When GPU memory is insufficient, you can proceed with training by adjusting the value of `--gradient_accumulation_steps`.

Also can use [Automatic Mixed Precision(Amp)](https://nvidia.github.io/apex/amp.html) to reduce memory usage and train faster

```

python3 train.py --name cifar10-100_500 --dataset cifar10 --model_type ViT-B_16 --pretrained_dir checkpoint/ViT-B_16.npz --fp16 --fp16_opt_level O2

```

## Results

To verify that the converted model weight is correct, we simply compare it with the author's experimental results. We trained using mixed precision, and `--fp16_opt_level` was set to O2.

### imagenet-21k

* [**tensorboard**](https://tensorboard.dev/experiment/Oz9GmmQIQCOEr4xbdr8O3Q)

|    model     |  dataset  | resolution | acc(official) | acc(this repo) |  time   |

|:------------:|:---------:|:----------:|:-------------:|:--------------:|:-------:|

|   ViT-B_16   | CIFAR-10  |  224x224   |       -       |     0.9908     | 3h 13m  |

|   ViT-B_16   | CIFAR-10  |  384x384   |    0.9903     |     0.9906     | 12h 25m |

|   ViT_B_16   | CIFAR-100 |  224x224   |       -       |     0.923      |  3h 9m  |

|   ViT_B_16   | CIFAR-100 |  384x384   |    0.9264     |     0.9228     | 12h 31m |

| R50-ViT-B_16 | CIFAR-10  |  224x224   |       -       |     0.9892     | 4h 23m  |

| R50-ViT-B_16 | CIFAR-10  |  384x384   |     0.99      |     0.9904     | 15h 40m |

| R50-ViT-B_16 | CIFAR-100 |  224x224   |       -       |     0.9231     | 4h 18m  |

| R50-ViT-B_16 | CIFAR-100 |  384x384   |    0.9231     |     0.9197     | 15h 53m |

|   ViT_L_32   | CIFAR-10  |  224x224   |       -       |     0.9903     | 2h 11m  |

|   ViT_L_32   | CIFAR-100 |  224x224   |       -       |     0.9276     |  2h 9m  |

|   ViT_H_14   | CIFAR-100 |  224x224   |       -       |      WIP       |         |

### imagenet-21k + imagenet2012

* [**tensorboard**](https://tensorboard.dev/experiment/CXOzjFRqTM6aLCk0jNXgAw/#scalars)

|    model     |  dataset  | resolution |  acc   |

|:------------:|:---------:|:----------:|:------:|

| ViT-B_16-224 | CIFAR-10  |  224x224   |  0.99  |

| ViT_B_16-224 | CIFAR-100 |  224x224   | 0.9245 |

|   ViT-L_32   | CIFAR-10  |  224x224   | 0.9903 |

|   ViT-L_32   | CIFAR-100 |  224x224   | 0.9285 |

### shorter train

* In the experiment below, we used a resolution size (224x224).

* [**tensorboard**](https://tensorboard.dev/experiment/lpknnMpHRT2qpVrSZi10Ag/#scalars)

|  upstream   |  model   |  dataset  | total_steps /warmup_steps | acc(official) | acc(this repo) |

|:-----------:|:--------:|:---------:|:-------------------------:|:-------------:|:--------------:|

| imagenet21k | ViT-B_16 | CIFAR-10  |          500/100          |    0.9859     |     0.9859     |

| imagenet21k | ViT-B_16 | CIFAR-10  |         1000/100          |    0.9886     |     0.9878     |

| imagenet21k | ViT-B_16 | CIFAR-100 |          500/100          |    0.8917     |     0.9072     |

| imagenet21k | ViT-B_16 | CIFAR-100 |         1000/100          |    0.9115     |     0.9216     |

## Visualization

The ViT consists of a Standard Transformer Encoder, and the encoder consists of Self-Attention and MLP module.

The attention map for the input image can be visualized through the attention score of self-attention.

Visualization code can be found at [visualize_attention_map](./visualize_attention_map.ipynb).

![fig3](./img/figure3.png)

## Reference

* [Google ViT](https://github.com/google-research/vision_transformer)

* [Pytorch Image Models(timm)](https://github.com/rwightman/pytorch-image-models)

## Citations

```bibtex

@article{dosovitskiy2020,

  title={An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale},

  author={Dosovitskiy, Alexey and Beyer, Lucas and Kolesnikov, Alexander and Weissenborn, Dirk and Zhai, Xiaohua and Unterthiner, Thomas and  Dehghani, Mostafa and Minderer, Matthias and Heigold, Georg and Gelly, Sylvain and Uszkoreit, Jakob and Houlsby, Neil},

  journal={arXiv preprint arXiv:2010.11929},

  year={2020}

}

```