README.md

OriNet-demo

This is the testing code for OriNet on MPI-INF-3DHP test set. Traing code will be released later.

The model was first pretrained on MPII and Human3.6M dataset and then fine-tuned on MPI-INF-3DHP training set without using any background augmentation.

The pretrained models can be downloaded here:

• Model trained on Human3.6m
• Model finetuned on MPI-INF-3DHP

Dependencies:

• Torch 7
• hdf5
• cudnn

Testing

• Download the MPI-INF-3DHP dataset and annotation file and put the h5 file in data/mpi/
• Suppose that the test sequences are located in mpi-inf-3dhp/test/
• In directory src, run th test.lua -dataDir /path/to/mpi-inf-3dhp -loadModel /path/to/your/model to save the results in Result.txt
• Go to directorytest_util and run evaluate.m in matlab to get the evaluation results. Remember to change the path to your annotation files. (The evaluation codes are provided in MPI-INF-3DHP dataset)

Results on MPI-INF-3DHP test set by activities

(Update) There has been a minor correction to the annotations for TS3 and TS4 in the test set. You should get the updated results based on this codebase.

	Stand/ Walk	Exercies	Sit on Chair	Crouch/ Reach	On the Floor	Sports	Misc.	All PCK	All AUC	MPJPE(mm)
Ours	95.5	82.3	89.9	84.6	66.5	92.0	93.0	84.3	47.5	84.5

Previous results ( for comparasion with published works)

	Stand/ Walk	Exercies	Sit on Chair	Crouch/ Reach	On the Floor	Sports	Misc.	All PCK	All AUC	MPJPE(mm)
VNect	87.7	77.4	74.7	72.9	51.3	83.3	80.1	76.7	40.4	124.7
Meta	86.6	75.3	74.8	73.7	52.2	82.1	77.5	75.7	39.3	117.6
Ours	90.4	79.1	88.5	81.6	66.3	91.9	92.2	81.8	45.2	89.4

Acknowledgements

We thank Alejandro Newell et al. for their great work and repo.

Reference

If you find our work useful in your research, please consider citing our paper:

@inproceedings{luo2018orinet,
  title     = {OriNet: A Fully Convolutional Network for 3D Human Pose Estimation}
  author    = {Luo, Chenxu and Chu, Xiao and Yuille, Alan}
  booktitle = {BMVC}
  year      = {2018}
}