Autodl复现Swin-Transformer

Swin_T是VMamba、MLLA模型的Baseline，在CV中是经典模型，它在目标检测任务（Object Detection）上有一个专属的名字（MMdetection），也有专属的库（mmdet）。Git网址如下：

GitHub - SwinTransformer/Swin-Transformer-Object-Detection: This is an official implementation for "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows" on Object Detection and Instance Segmentation.This is an official implementation for "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows" on Object Detection and Instance Segmentation. - SwinTransformer/Swin-Transformer-Object-Detectionhttps://github.com/SwinTransformer/Swin-Transformer-Object-Detection.git本文主要使用Swin的与训练好的模型，实现COCO数据集的验证

基本环境

镜像选择

coco数据下载

使用autodl的云端（足够快,推荐）

sudo apt-get install unzip  
mkdir -p /root/autodl-tmp/coco/
unzip /root/autodl-pub/COCO2017/annotations_trainval2017.zip -d /root/autodl-tmp/coco/
unzip /root/autodl-pub/COCO2017/val2017.zip -d /root/autodl-tmp/coco/
unzip /root/autodl-pub/COCO2017/train2017.zip -d /root/autodl-tmp/coco/
unzip /root/autodl-pub/COCO2017/test2017.zip -d /root/autodl-tmp/coco/

环境设置 

pip install mmcv-full==1.3.9 -f https://download.openmmlab.com/mmcv/dist/cu111/torch1.9.0/index.html
pip install -U openmim
mim install mmengine
​​​​​​​pip install timm
​​​​​​​pip install Cython==0.29.33
pip install apex





 

克隆Swin-T

cd
git clone https://github.com/SwinTransformer/Swin-Transformer-Object-Detection.git
cd Swin-Transformer-Object-Detection
pip install -e.
pip install yapf==0.31

下载预训练数据

cd
wget https://github.com/SwinTransformer/storage/releases/download/v1.0.3/mask_rcnn_swin_tiny_patch4_window7_1x.pth

测试部分

修改路径

把Swin-Transformer-Object-Detection/tools/test.py得test.py文件进行插入自己的路径

 插入代码如下

# 修改数据路径
new_data_root = '/root/autodl-tmp/coco/'

# 修改 data_root
cfg.data_root = new_data_root

# 更新数据加载器中引用的路径
if hasattr(cfg.data, 'train'):
    cfg.data.train.ann_file = cfg.data.train.ann_file.replace('data/coco/', new_data_root)
    cfg.data.train.img_prefix = cfg.data.train.img_prefix.replace('data/coco/', new_data_root)

if hasattr(cfg.data, 'val'):
    cfg.data.val.ann_file = cfg.data.val.ann_file.replace('data/coco/', new_data_root)
    cfg.data.val.img_prefix = cfg.data.val.img_prefix.replace('data/coco/', new_data_root)

if hasattr(cfg.data, 'test'):
    cfg.data.test.ann_file = cfg.data.test.ann_file.replace('data/coco/', new_data_root)
    cfg.data.test.img_prefix = cfg.data.test.img_prefix.replace('data/coco/', new_data_root)

开始测试

python /root/Swin-Transformer-Object-Detection/tools/test.py /root/Swin-Transformer-Object-Detection/configs/swin/mask_rcnn_swin_tiny_patch4_window7_mstrain_480-800_adamw_1x_coco.py /root/mask_rcnn_swin_tiny_patch4_window7_1x.pth  --eval bbox segm

训练部分

把Swin-Transformer-Object-Detection/tools/train.py得train.py文件进行插入自己的路径

 代码如下

new_data_root = '/root/autodl-tmp/coco/'

# 更新数据集路径中的相关部分
cfg.data_root = new_data_root

# 训练集、验证集和测试集路径
if 'train' in cfg.data:
    cfg.data['train']['img_prefix'] = cfg.data['train']['img_prefix'].replace('data/coco/', new_data_root)
    cfg.data['train']['ann_file'] = cfg.data['train']['ann_file'].replace('data/coco/', new_data_root)

if 'val' in cfg.data:
    cfg.data['val']['img_prefix'] = cfg.data['val']['img_prefix'].replace('data/coco/', new_data_root)
    cfg.data['val']['ann_file'] = cfg.data['val']['ann_file'].replace('data/coco/', new_data_root)

if 'test' in cfg.data:
    cfg.data['test']['img_prefix'] = cfg.data['test']['img_prefix'].replace('data/coco/', new_data_root)
    cfg.data['test']['ann_file'] = cfg.data['test']['ann_file'].replace('data/coco/', new_data_root)

# 更新评估器路径中的 ann_file
if 'val_evaluator' in cfg and 'ann_file' in cfg.val_evaluator:
    cfg.val_evaluator['ann_file'] = cfg.val_evaluator['ann_file'].replace('data/coco/', new_data_root)

if 'test_evaluator' in cfg and 'ann_file' in cfg.test_evaluator:
    cfg.test_evaluator['ann_file'] = cfg.test_evaluator['ann_file'].replace('data/coco/', new_data_root)

if 'train_evaluator' in cfg and 'ann_file' in cfg.train_evaluator:
    cfg.train_evaluator['ann_file'] = cfg.train_evaluator['ann_file'].replace('data/coco/', new_data_root)

开始训练

python /root/Swin-Transformer-Object-Detection/tools/train.py /root/Swin-Transformer-Object-Detection/configs/swin/mask_rcnn_swin_tiny_patch4_window7_mstrain_480-800_adamw_1x_coco.py