Best of Both Worlds: See and Understand Clearly in the Dark (ACM MM2022)

The architecture and training stategy

Codes

Requirements

• Python 3.6
• Pytorch==1.14.0
• Cuda 11.1

How to use

• For training
  The traning codes will be released with the journal version.
• For testing
  For low-light image enhancement, run test_en.py with enhance.pth.
  For dark face detection/nighttime segmentation, run test_det.py. Trained detection model is in https://drive.google.com/file/d/10zb5uC7j0N7fspG2HXljKQEYhRQUUaTQ/view?usp=drive_link.
  For nighttime semantic segmentation, visual results (ACDC night val set) of trained segmentation model are in https://drive.google.com/drive/folders/11qfyxLXjH_sQF1t4-4sRCEzHAZ0BywgJ?usp=drive_link.