Joint Person Segmentation and Identification in Synchronized First- and Third-person Videos

Introduction

This is a PyTorch implementation for our ECCV 2018 paper "Joint Person Segmentation and Identification in Synchronized First- and Third-person Videos".

Environment

The code is developed with CUDA 8.0, Python 2.7, PyTorch 0.3

Data Preparation

• Download IU ShareView dataset and pretrained models at http://vision.soic.indiana.edu/firstthird-eccv2018.

Citations

If you are using the data/code/model provided here in a publication, please cite our paper:

@inproceedings{firstthirdseg2018eccv,
    title = {Joint Person Segmentation and Identification in Synchronized First- and Third-person Videos},
    author = {Mingze Xu and Chenyou Fan and Yuchen Wang and Michael S. Ryoo and David J. Crandall},
    booktitle = {European Conference on Computer Vision (ECCV)},
    year = {2018}
}