This repository contains the code for the CVPR 2020 paper "Differentiable Volumetric Rendering: Learning Implicit 3D Representations without 3D Supervision"
When I train the single_view_reconstruction with multi-view supervision model, during the evaluation, there is a key error: loss_depth_eval. I found that it was due to the predicted mask was zero. What could be the problem?
When I train the single_view_reconstruction with multi-view supervision model, during the evaluation, there is a key error:
loss_depth_eval
. I found that it was due to the predicted mask was zero. What could be the problem?