Home

Awesome

Graph Neural Based End-to-end Data Association Framework for Online Multiple-Object Tracking

A PyTorch implementation combines with Siamese Network and Graph Neural Network for Online Multiple-Object Tracking.

Dataset available at [https://motchallenge.net/]

According paper can be found at [https://arxiv.org/abs/1907.05315]

How to run

Use python main.py to train a model from scratch. Settings for training is in config.yml.
Use python tracking.py to track a test video, meanwhile you need to provide the detected objects & tracking results for the first five frames. Setting for tracking is in setting/.

Requirements