mtlong / pose_estimation

Code for our work on pose-estimation using template 3D models.

Template-Model Driven Pose Estimation

This repository contains the code for our works:

1. iSPA-NET, accepted in ACM Multimedia Conference, 2018 (ACMMM18) - *Jogendra Nath Kundu, *Aditya Ganeshan, *Rahul MV, Aditya Prakash, R. Venkatesh Babu

2. Object Pose Estimation Using Multi-View Keypoint Correspondence, accepted in Geometry Meets Deep Learning Workshop, at ECCV 2018. - *Jogendra Nath Kundu, *Rahul MV, *Aditya Ganeshan, R. Venkatesh Babu

* = equal Contribution

This repository broadly consists of:

1. Data Creation : data_creator

2. Universal Correspondence Network Training Code : ucn_code

3. iSPA-Net Training and Testing code : classifier_code

   • Refer to classifier_code/iSPA-Net.md for more details

4. Multi-View Keypoint based Pose Estimation - Training and Testing code : classifier_code

   • Refer to classifier_code/Multi-View-Net.md for more details

Note: For the Universal Correspondence Network, caffe code for which has to be requested from NEC-Labs.

iSPA-Net Link to Arxiv: 1808.01134

Object Pose Estimation Using Multi-View Keypoint Correspondence - 1809.00553

Citation

#@article{ispanet-kundu-2018,
title={iSPA-Net: Iterative Semantic Pose Alignment Network},
author={Kundu, Jogendra Nath and Ganeshan, Aditya and MV, Rahul and Prakash, Aditya and Babu, R Venkatesh},
booktitle = {arXiv preprint arXiv: 1808.01134 },
year = {2018}
}

#@article{ispanet-kundu-2018,
title={Object Pose Estimation Using Multi-View Keypoint Correspondence},
author={Kundu, Jogendra Nath and MV, Rahul and Ganeshan, Aditya and Babu, R Venkatesh},
booktitle = {arXiv preprint arXiv: 1809.00553 },
year = {2018}
}