1. Set up conda environment via

conda env create --file env.yaml

2. Download the raindrop image from Onedrive and place it under the directory './icdar2015_rain'. It consists of at least two folders: 'ch4_test_images' and 'ch4_test_images_gt'
3. Download pre-trained checkpoint for ICDAR 2015 dataset: LINK
4. Running command for inference is as below. If the argument --derain is prepended, the image is processed by the de-rain module before the text detection module is running.

python eval.py --trained_model=./craft_ic15_20k.pth --test_folder=./icdar2015_rain [--derain]

ToDo

• Training code: we need to clarify which ablation studies will be conducted.