search

ying09 / TextFuseNet

A PyTorch implementation of "TextFuseNet: Scene Text Detection with Richer Fused Features".
MIT License
475 stars 122 forks source link

About icdar2015 dataset #76

Closed chenf99 closed 2 years ago

chenf99 commented 3 years ago

Hello，I want to train TextFuseNet on icdar2015 dataset, it seems that I should convert the txt of train/test set to json. However, in configs/ocr/icdar2015_101_FPN.yaml, both the TRAIN dataset and TEST dataset are icdar2015. According to detectron2/data/datasets/builtin.py, the json of "icdar2015" is icdar2015/train.json. So, is the yaml file wrong？Why the train set and the test set are the same?

Real-YeJ commented 3 years ago

@chenf99 When testing, please run the code in demo (icdar2015_detection.py), and the detection results of test images will be saved as txt format. For evaluation, please refer to the link of icdar2015 in Evaluation part.