Giters

KunalMGupta / NeuralMeshFlow

Neural Mesh Flow: 3D Manifold Mesh Generation via Diffeomorphic Flows

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

ShapeNet Dataset

low5545 opened this issue · comments

commented

In Section 4 (Data) of the manuscript, you mentioned that the training and testing splits used are in accordance to 3D-R2N2. However upon inspection of the pre-processed point cloud samples and normals, I realized that the number of instances in most of the shape categories do not tally with the image renders from 3D-R2N2, especially for the car (02958343) synset. I am aware that 3D-R2N2 uses ShapeNetCore_v1, am I right that the point cloud samples and normals in this work are generated from ShapeNetCore_v2? Thank you.