xy-guo / Learning-Monocular-Depth-by-Stereo

Learning Monocular Depth by Distilling Cross-domain Stereo Networks, ECCV18
https://arxiv.org/abs/1808.06586
MIT License
94 stars 8 forks source link

About The Dataset #2

Closed jinjingyi closed 5 years ago

jinjingyi commented 5 years ago

Thanks for your work. For the preparation of the dataset, can you give the relevant links and provide more detailed instructions?

xy-guo commented 5 years ago

You can first refer to the Monodepth GitHub repo for the preparation of KITTI dataset. For the Scene Flow datasets, you can first download all the tars (Finalpass and disparities of monkaa, flyingthings, driving) and then extract and place the folders according to the sceneflow_folder_structure.txt. For automatical script, I may update next week.

jinjingyi commented 5 years ago

Thanks for your response, I will try it as you said and look forward to your update.

xy-guo commented 5 years ago

README has been updated.