《Dual Aggregation Transformer for Image Super-Resolution》（DAT）-- ICCV 2023 代码部署 kaggle

最近实验室服务器用不了，看论文复现测试代码只能找这些在线平台，记录一下自己部署github项目到kaggle的过程。

github：GitHub - zhengchen1999/DAT: PyTorch code for our ICCV 2023 paper "Dual Aggregation Transformer for Image Super-Resolution".

环境配置：

Python 3.8

Pytorch 1.8.0

NVIDIA GPU + CUDA

# Clone the github repo and go to the default directory 'DAT'.
git clone https://github.com/zhengchen1999/DAT.git
conda create -n DAT python=3.8
conda activate DAT
pip install -r requirements.txt
python setup.py develop

数据集：https://drive.google.com/drive/folders/1ZMaZyCer44ZX6tdcDmjIrc_hSsKoMKg2?usp=drive_link

预训练模型：https://drive.google.com/drive/folders/1iBdf_-LVZuz_PAbFtuxSKd_11RL1YKxM?usp=drive_link

作者提供的可视化结果：https://drive.google.com/drive/folders/1ZMaZyCer44ZX6tdcDmjIrc_hSsKoMKg2?usp=drive_link

测试：

# Test on your dataset
python basicsr/test.py -opt options/Test/test_single_x2.yml
python basicsr/test.py -opt options/Test/test_single_x3.yml
python basicsr/test.py -opt options/Test/test_single_x4.yml

kaggle部署DAT项目并进行测试：

! git clone https://github.com/zhengchen1999/DAT.git
! pip install -r /kaggle/working/DAT/requirements.txt
​
!pip install tqdm
!pip install yapf
!pip install timm
!pip install einops
!pip install h5py
​
!pip3 install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu118
​
!gdown --folder https://drive.google.com/drive/folders/14VG5mw5ie8RrR4jjypeHynXDZYWL8w-r?usp=sharing -O /kaggle/working/DAT/experiments/pretrained_models/
​
!cp -r /kaggle/input/dat-datasets-test /kaggle/working/DAT/datasets
​
cd /kaggle/working/DAT
​
%%writefile /kaggle/working/DAT/options/Test/test_DAT_x4.yml
# general settings
name: test_DAT_x4
model_type: DATModel
scale: 4
num_gpu: 2
manual_seed: 10
​
datasets:
  test_1:  # the 1st test dataset
    task: SR
    name: Set5
    type: PairedImageDataset
    dataroot_gt: /kaggle/working/DAT/datasets/dat-datasets-test/benchmark/Set5/HR
    dataroot_lq: /kaggle/working/DAT/datasets/dat-datasets-test/benchmark/Set5/LR_bicubic/X4
    filename_tmpl: '{}x4'
    io_backend:
      type: disk
​
  test_2:  # the 2st test dataset
    task: SR
    name: Set14
    type: PairedImageDataset
    dataroot_gt: /kaggle/working/DAT/datasets/dat-datasets-test/benchmark/Set14/HR
    dataroot_lq: /kaggle/working/DAT/datasets/dat-datasets-test/benchmark/Set14/LR_bicubic/X4
    filename_tmpl: '{}x4'
    io_backend:
      type: disk
​
  test_3:  # the 3st test dataset
    task: SR
    name: B100
    type: PairedImageDataset
    dataroot_gt: /kaggle/working/DAT/datasets/dat-datasets-test/benchmark/B100/HR
    dataroot_lq: /kaggle/working/DAT/datasets/dat-datasets-test/benchmark/B100/LR_bicubic/X4
    filename_tmpl: '{}x4'
    io_backend:
      type: disk
​
  test_4:  # the 4st test dataset
    task: SR
    name: Urban100
    type: PairedImageDataset
    dataroot_gt: /kaggle/working/DAT/datasets/dat-datasets-test/benchmark/Urban100/HR
    dataroot_lq: /kaggle/working/DAT/datasets/dat-datasets-test/benchmark/Urban100/LR_bicubic/X4
    filename_tmpl: '{}x4'
    io_backend:
      type: disk
​
  test_5:  # the 5st test dataset
    task: SR
    name: Manga109
    type: PairedImageDataset
    dataroot_gt: /kaggle/working/DAT/datasets/dat-datasets-test/benchmark/Manga109/HR
    dataroot_lq: /kaggle/working/DAT/datasets/dat-datasets-test/benchmark/Manga109/LR_bicubic/X4
    filename_tmpl: '{}_LRBI_x4'
    io_backend:
      type: disk
​
​
# network structures
network_g:
  type: DAT
  upscale: 4
  in_chans: 3
  img_size: 64
  img_range: 1.
  split_size: [8,32]
  depth: [6,6,6,6,6,6]
  embed_dim: 180
  num_heads: [6,6,6,6,6,6]
  expansion_factor: 4
  resi_connection: '1conv'
​
# path
path:
  pretrain_network_g: /kaggle/working/DAT/experiments/pretrained_models/DAT/DAT_x4.pth
  strict_load_g: True
​
# validation settings
val:
  save_img: True
  suffix: ~  # add suffix to saved images, if None, use exp name
  use_chop: False  # True to save memory, if img too large
​
  metrics:
    psnr: # metric name, can be arbitrary
      type: calculate_psnr
      crop_border: 4
      test_y_channel: True
    ssim:
      type: calculate_ssim
      crop_border: 4
      test_y_channel: True
​
​
!python /kaggle/working/DAT/basicsr/test.py -opt /kaggle/working/DAT/options/Test/test_DAT_x4.yml