This is the codebase for our technical report "Improved Techniques for Learning to Dehaze and Beyond: A Collective Study"
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.

Improved Techniques for Learning to Dehaze and Beyond: A Collective Study


This is the official codebase for our paper "Improved Techniques for Learning to Dehaze and Beyond: A Collective Study".

The paper reviews the collective endeavors by the team of authors in exploring two interlinked important tasks, based on the recently released REalistic Single Image DEhazing (RESIDE) benchmark: i) single image dehazing as a low-level image restoration problem; ii) high-level visual understanding (e.g., object detection) from hazy images. For the first task, the authors investigated on a variety of loss functions, and found perception-driven loss to improve dehazing performance very notably. For the second task, the authors came up with multiple solutions including using more advanced modules in the dehazing-detection cascade, as well as domain-adaptive object detectors. In both tasks, our proposed solutions are verified to significantly advance the state-of-the-art performance.

Code organization

Each individual software package and corresponding documentation are located under code/PACKAGE_NAME


See code/pad_net

Domain adaptation for MaskRNN

See code/adapt_maskrnn

Improving Object Detection in Haze

See code/iodh

Sandeep and Satya's work

see code/sandeep_satya


This collective study was initially performed as a team project effort in the Machine Learning course (CSCE 633, Spring 2018) of CSE@TAMU, taught by Dr. Zhangyang Wang. We acknowledge the Texas A&M High Performance Research Computing (HPRC) for providing a part of the computing resources used in this research.



  title={Improved Techniques for Learning to Dehaze and Beyond: A Collective Studys},
  author={Yu Liu, Guanlong Zhao, Boyuan Gong, Yang Li, Ritu Raj, Niraj Goel, Satya Kesav, Sandeep Gottimukkala, Zhangyang Wang, Wenqi Ren, Dacheng Tao},