Home

Awesome

flownet3d_pytorch

The pytorch implementation of flownet3d based on WangYueFt/dcp, sshaoshuai/Pointnet2.PyTorch and yanx27/Pointnet_Pointnet2_pytorch

Installation

Requirements

PyTorch>=1.0: https://pytorch.org

scipy>=1.2

numpy

h5py

tqdm

Install

Install this library by running the following command:

cd lib
python setup.py install
cd ../

Training

The processed Flyingthings3d data is provided here for download (total size ~11GB).

Then run the following command to train:

python main.py --exp_name=flownet3d --dataset_path=xx/yy

where xx/yy is the dataset path

Performance comparison

All of the following experiments were tested on a TITAN RTX

  1. GPU memory usage:
batch sizeflownet3d(GB)flownet3d_pytorch(GB)
1616.99.1
  1. Training time per epoch on Flyingthings3d dataset:
batch sizeflownet3d(min)flownet3d_pytorch(min)
166.73.4