Figure 3From: Weakly-Supervised Single-view Dense 3D Point Cloud Reconstruction via Differentiable RendererThe GT depth value is sampled by bilinear interpolationBack to article page