southmelon22 / east_resnet50 Goto Github PK

This is a PyTorch Re-Implementation of EAST: An Efficient and Accurate Scene Text Detector.

• ResNet50 as the feature extractor backbone.
• Feature Pyramid Network in the feature merging branch.
• ReduceLROnPlateau as the lr scheduler with 0.8 factor and 2000 patience.
• Only RBOX part is implemented.
• The pre-trained model provided achieves 82.12 F-score on ICDAR 2015 Challenge 4 using only the 1000 images.

Required libraries

• Anaconda 3
• PyTorch 1.10.1
• Lanms 1.0.2
• Shapely 1.8.0
• Opencv-python 4.5.4.60
• Pip 21.2.4
• Python 3.9.7
• Scipy 1.8.0
• Numpy 1.20.1
• Wandb
• Pillow 9.0.1
• Torchvision 0.11.2
• Importlib 4.8.1
• Matplotlib 3.4.3

git clone https://github.com/ChunFei96/EAST_resnet50.git
cd EAST_resnet50

• Download Train and Test Data: ICDAR 2015 Challenge 4. Cut the data into four parts: train_img, train_gt, test_img, test_gt.

.
├── EAST_resnet50
│   ├── evaluate
│   ├── lanms
│   ├── pths
│   ├── torchvision
|   ├── dataset.py
|   ├── detect.py
|   ├── eval.py
|   ├── loss.py
|   ├── model.py
|   ├── resnet.py
|   ├── train.py
|   └── utils.py
└── ICDAR_2015
    ├── test_gt
    ├── test_img
    ├── train_gt
    └── train_img

Modify the parameters in train.py and run:

CUDA_VISIBLE_DEVICES=0,1 python train.py

Modify the parameters in detect.py and run:

CUDA_VISIBLE_DEVICES=0 python detect.py

• The evaluation scripts are from ICDAR Offline evaluation and have been modified to run successfully with Python 3.7.1.
• Change the evaluate/gt.zip if you test on other datasets.
• Modify the parameters in eval.py and run:

CUDA_VISIBLE_DEVICES=0 python eval.py

• wandb.ai is used to record and monitor live metrics such as Recall, Precision and F1-score.
• Please modify the parameter under the wandb init and config to direct to your wandb project name and entity in train.py.

wandb.init(project="xxx", entity="xxx")

wandb.config = {
        "pths_path": pths_path,
        "batch_size": batch_size,
        "lr": lr,
        "num_workers": num_workers,
        "epoch_iter": epoch_iter,
        "save_interval": save_interval,
    }