Each pixel value of the depth map is 0.01

TRI-ML / packnet-sfm

TRI-ML Monocular Depth Estimation Repository

https://tri-ml.github.io/packnet-sfm/

MIT License

1.23k stars 242 forks source link

Each pixel value of the depth map is 0.01 #180

Closed mengjiaos closed 2 years ago

mengjiaos commented 2 years ago

When I use NRS to train on the omnicam dataset, when inferring, the each pixel's value of the depth map is 0.01, and a yellow depth map is displayed. 000000000

VitorGuizilini-TRI commented 2 years ago

Seems like your depth maps is not converging, can you please try a smaller learning rate?

mengjiaos commented 2 years ago

@VitorGuizilini-TRI Thanks, I have solve the problem by modifying learning rate to 0.00009.