Open 120829 opened 4 months ago
Could you show some results of how it looks? I think that would help (I am not the author of this repo, but I work in a similar space).
Could you show some results of how it looks? I think that would help (I am not the author of this repo, but I work in a similar space).
I have bad results when depth supervision involved, every train camera view is converged well and the render image looks good, however, the test views are bad. So I check the point cloud and I find that the point cloud converged in the front of every train pose, but they don't compose to our object, I think it's the depth shift and scale problem, can you help me with this question, Thank you!
Could you show some results of how it looks? I think that would help (I am not the author of this repo, but I work in a similar space).
I have three train pose, the point cloud gathered in front of them, but they don't compose the object. Under every train view the point cloud project a good shape, this is a hornbill by the way:
I followed your steps to run the code. Why did the performance deteriorate after adding the depth map?