Skip to content

Latest commit

 

History

History
35 lines (19 loc) · 1.03 KB

File metadata and controls

35 lines (19 loc) · 1.03 KB

Color Universal Design Neural Network for the Color Vision Deficiencies

This repository is the official implementation of Color Universal Design Neural Network for the Color Vision Deficiencies

Notice

  • As our train dataset is exclusive to cooperation institutes, only part of the inference dataset can be released.
  • The HSV conversion algorithm has malfunction of reversing Hue on the value with 180.
  • We used extra training data on up-loaded model, so there would be some error comparing to papers'

Requirements

  • torch >= 1.5.1

Training and Evaluation

python main.py --log_interval 100 --save_interval 10 -cuda -pin_memory --mode train --worker 8
python test_image.py -cuda -pin_memory --worker 2 --mode inference --data_class A

Pre-trained Models

You can find pretrained models on "model_checkpoints/":

Results

image_1

License

License: MIT