OpenStereo: A Comprehensive Benchmark for Stereo Matching and Strong Baseline
465
stars
51
forks
source link
使用kitti数据集训练psmnet时似乎有问题,使用train.py训练时打印出的eval数据集的epe过低(一点几,和loss相近),使用eval.py打印出的eval数据集的epe过高(三十几),这与许多论文的结果不一致 #101
Closed
XianghuiFan closed 5 months ago
Please check the cfgs/psmnet/psmnet_kitti.yaml file in v2 branch.