Skip to content

liufeng2915/Implicit_Dense_Correspondence

Repository files navigation

Learning Implicit Functions for Topology-Varying Dense 3D Shape Correspondence

Advances in Neural Information Processing Systems (NeurIPS 2020). Oral presentation. [Arxiv, PDF, Supp, Project]

Feng Liu, Xiaoming Liu

Department of Computer Science and Engineering, Michigan State University

The goal of this paper is to learn dense 3D shape correspondence for topology-varying objects in an unsupervised manner.

teaser

Figure 1: Given a shape S, PointNet E is used to extract the shape feature code z. Then a part embedding o is produced via a deep implicit function f. We implement dense correspondence through an inverse function mapping from o to recover the 3D shape. (b) To further make the learned part embedding consistent across all the shapes, we randomly select two shapes SA and SB. By swapping the part embedding vectors, a cross reconstruction loss is used to enforce the inverse function to recover to each other.


This code is developed with Python3 and PyTorch 1.1

Dataset

Please refer to data/README

Pretrained models

Please refer to models/README

Evaluation

Please refer to evaluation/README

Training

Our method is trained in three stages: 1) PointNet like encoder and implicit function are trained on sampled point-value pairs via occupancy loss. 2) encoder, implicit function and inverse implicit function are jointly trained via occupancy and self-reconstruction losses. 3) We jointly train encoder, implicit function and inverse implicit function with occupancy, self-reconstruction and cross-reconstruction losses.

## ----------------------  chair  ----------------------------- ##
# stage1
CUDA_VISIBLE_DEVICES=0 python train_stage1.py --cate_name chair --epoch  50 --resolution 16 --batch_size 45 
CUDA_VISIBLE_DEVICES=0 python train_stage1.py --cate_name chair --epoch 100 --resolution 32 --batch_size 25 --pretrain_model TRUE --pretrain_model_name Corr-49.pth
CUDA_VISIBLE_DEVICES=0 python train_stage1.py --cate_name chair --epoch 300 --resolution 64 --batch_size  7 --pretrain_model TRUE --pretrain_model_name Corr-99.pth

# stage2
CUDA_VISIBLE_DEVICES=0 python train_stage2.py --cate_name chair --epoch 500 --resolution 64 --batch_size  4 --pretrain_model TRUE --pretrain_model_name stage1/Corr-299.pth

# stage3
CUDA_VISIBLE_DEVICES=0 python train_stage3.py --cate_name chair --epoch 1000 --resolution 64 --batch_size  2 --pretrain_model TRUE --pretrain_model_name stage2/Corr-499.pth

Contact

For questions feel free to post here or directly contact the author via liufeng6@msu.edu

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published