search

microsoft / DIF-Net

Deformed Implicit Field: Modeling 3D Shapes with Learned Dense Correspondence CVPR 2021
MIT License
121 stars 19 forks source link

ShapeNet dataset #6

Open 2j1ejyu opened 2 years ago

2j1ejyu commented 2 years ago

Hello. Thank you for your work. I'd like to ask how you've chosen data for training or evaluation. For example, ShapeNetCore.v2 consists of about 6000 chair objects. However, only 4100 objects were used for training and evaluating DIF-Net.

YuDeng commented 2 years ago

Hi, for training, we simply choose the first 4000 objects sorted by the object name (md5 file); for evaluation, we first calculate the intersections between the remaining objects and the evaluation set used by DeepSDF, and then choose the first 100 objects sorted by md5 name. Objects used can be found in ./split folder.

2j1ejyu commented 2 years ago

Thank you very much for your quick response!