Pretrained network to predict depth map

[zhangwenniu]

Hi, thanks for your great job creating and calibrating transparent object dataset.
Since the paper claim that your dataset can facilitate depth map filling works, can you share the pretrained model?
I would like to predict depth map of transparent objects in real captured scene, it seems that train a network on your dataset from scratch is resource-consuming and maybe not able to reproduce the same results with yours due to experiment configuration. I would appreciate that you could share a pretrained model trained on Clear-Pose dataset to predict depth map or normal maps.
Thanks for your help and congratulations to your great jobs.