facebookresearch / DistDepth

Repository for "Toward Practical Monocular Indoor Depth Estimation" (CVPR 2022)
Other
216 stars 20 forks source link

Hello, I used the smaller copy set "SimSIN-simple" provided by you for training, and then I ran "demo.py" with the obtained weights, hoping to get some expected results, but the results were not good, I felt completely ineffective, and the pictures I got were all purple, may I ask why I did not modify your code #26

Open iSuryxin opened 1 year ago

2579690686 commented 11 months ago

I'm having the same issue, do you have any results?

2579690686 commented 11 months ago

If you look at the loss during training, I think the loss must not converge

choyingw commented 11 months ago

see #25