CS231_Low-light Enhancement in Binary Image Classification

The low light enhancement for the Binary Image Classification Tasks project

I. Low-light enhancement techniques

II. Guide

Struture

Dataset

• Dowload the Exdark dataset (including images and annotations) in https://github.com/cs-chan/Exclusively-Dark-Image-Dataset.

• At dataset.py, replace in ExDark_pytorch.init the anno_dir and img_dir variables corresponding to your local path.

      class ExDark_pytorch(Dataset):
          def __init__(self, annotations_file, 
                      transform, 
                      enhance_type=None,
                      anno_dir=, "path_to_folder_of_images"
                      img_dir="path_to_folder of annator"): # 
              
              with open(annotations_file, "r") as f:
                  lines = f.readlines()
                  lines = [line.strip() for line in lines]
                  ....

Training and Testing

• Run following cmd to training and testing the image classification, the more detailed is in flag descriptions of train.py and test.py.

  python train.py --train_model classification --pretrained Trained_model/best_classify_NN.pth --train_annotator Splits/Train.txt --test_annotator Splits/Test.txt --enhanced_type log_transform

  python test.py --test_model classification --pretrained Trained_model/best_classify_NN.pth --test_annotator Splits/Test.txt --enhanced_type log_transform 

Requiremnts

pytorch cudn
tensorflow cudn