TD with DCCAM

Get Started

1. install required packages: including: Pytorch version 1.9.0, torchvision version 0.10.0 and Timm version 0.5.4

   pip install -r requirements.txt

   For mixed-precision training, please install apex

   For object detection, please additionally install detectron2 library and shapely. Refer to the Detectron2's INSTALL.md.

2. Prepare dataset available ICDAR2019-cTDaR

3. Download pre-trained weights. here

Train

Training in ICDAR 2019 cTDaR modern subset:

python train_net.py 

Detectron2's document may help you for more details.

Inference

One can run inference using the inference.py script.

python inference.py 

Acknowledgment

Thanks to DIT for providing the pre-trained weight framework.

Thanks to Detectron2 for Cascade Mask R-CNN implementation.

Thanks to YOLOv8 and ACFPN for providing the open-source framework.