LIFE: Lighting Invariant Flow Estimation
Zhaoyang Huang*, Xiaokun Pan*, Runsen Xu, Yan Xu, Ka Chun Cheung, Guofeng Zhang, Hongsheng Li
conda env create -f environment.yaml
conda activate life
- For single image warping, please refer to the
notebooks/flow_warp.ipynb. - LIFE can robustly align a picture to an image under significant lighting variation. To synthesize an image with realistic lighting conditions, please refer to our previous work NIID-Net.
- To play a movie on the paper, please refer to DEMO.md
Comming soon.
cd evaluation
python kitti_eval.py --model PATH_To_MODEL --version 2012/2015 --data_dir PATH_TO_KITTI_DATASET
Our neural network architecture is mainly built upon RAFT.