/camera_lens_glare

camera_lens_glare_reduction

Primary LanguageJupyter NotebookMIT LicenseMIT

camera_lens_glare

camera_lens_glare_reduction

-> log files saved in work_dir -> saved image files saved in reuslt_dir -> submission files saved in submission_dir

  1. Install prerequested files pip install -r requirements.txt

2.(Option) Install Apex https://github.com/NVIDIA/apex

*Install git clone https://github.com/NVIDIA/apex cd apex pip install -v --disable-pip-version-check --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" ./

#Python only build pip install -v --disable-pip-version-check --no-cache-dir ./

  1. Patch generation: python src/dataset/generate_patches.py --src_dir/dataset_sub/camera_light_glare/train/ --tar_dir /dataset_sub/camera_light_glare/patches_1024_sample/train_patch/ --ps 256 --num_patches 20

  2. Train CUDA_VISIBLE_DEVICES=0 python src/train_no_apex_refactor.py --batch_size 4 --max_epoch 50 --data_path /home/kt05/work/dataset/camera_light_glare/ --width 256 --height 256 --model_type resnet_unet --result_dir ./result_mirnet_no_pre/ --work_dir ./cur/deglare_mirnet_no_pre/ --submission_dir ./submission_mirnet/