Pose Baseline 3D PyTorch

A PyTorch reimplementation of the paper - 3D human pose estimation.

Julieta Martinez, Rayat Hossain, Javier Romero, James J. Little. A simple yet effective baseline for 3d human pose estimation. In ICCV, 2017. PDF: https://arxiv.org/pdf/1705.03098.pdf

You can check original Tensorflow implementation written by Julieta Martinez et al.

This repo reproduces the work by @weigq.

Dependencies

• PyTorch >= 1.0.0

Datasets

Human3.6m

Installation

git clone https://github.com/jaroslaw1007/Pose_Baseline_3D_PyTorch.git

Usage

Train

Train on Human3.6m groundtruth 2d joints.

python main.py --training --max_norm

Test

python main.py

	direct.	discuss.	eat.	greet.	phone	photo	pose	purch.	sit	sitd.	somke	wait	walkd.	walk	walkT	avg
Julieta Martinez et al.	37.7	44.4	40.3	42.1	48.2	54.9	44.4	42.1	54.6	58.0	45.1	46.4	47.6	36.4	40.4	45.5
weigq	35.7	42.3	39.7	40.7	44.5	53.3	42.8	40.1	52.5	53.9	42.8	43.1	44.1	33.4	36.3	43.0
This version	35.5	41.7	39.0	40.4	44.4	52.4	42.7	38.2	53.6	54.6	42.6	42.8	44.1	33.9	36.9	42.8

Citing

@inproceedings{martinez_2017_3dbaseline,
  title={A simple yet effective baseline for 3d human pose estimation},
  author={Martinez, Julieta and Hossain, Rayat and Romero, Javier and Little, James J.},
  booktitle={ICCV},
  year={2017}
}