LightStereo不能复现Pretrained Model的性能

[lihyin]

LightStereo-M-KITTI.ckpt的结果:

用这个配置https://github.com/lihyin/OpenStereo/blob/v2/cfgs/lightstereo/lightstereo_m_kitti.yaml重新训练490 Epoch的结果：

请问你们是怎么训练LightStereo-M-KITTI.ckpt的（哪个配置文件、多少Epoch）？

[XiandaGuo]

For the KITTI fine-tuning experiments, we pre-trained on the SceneFlow dataset using different data augmentation strategies. We will update the corresponding YAML configuration files and checkpoints as soon as possible.

[lihyin]

Thank you so much @XiandaGuo! Will you be able to update them in the next 3 days? Depending on how soon you can update, we can plan our work accordingly