/2019-CVPR-AIC-Track-3-UWIPL

Repository for 2019 CVPR AI City Challenge Track 3 from IPL@UW

Primary LanguagePython

2019 CVPR AI City Challenge Track 3

This repository is used to do traffic anomaly detection, including lane violation, traffic accident, emergency stop on highway, etc. Our method ranks 2nd in 2019 CVPR AI City Challenge Track 3.

Updates

For Collaborators: Please write your summary of your updates in the following format:
5/31/2019 Update all the code for time decision.
4/9/2019 Update scripts for candidate selection and models

Introduction

This is an overall description of the repository. Detailed descriptions could be found under each folders.
Under candidate selection are the scripts for getting anomaly candidate video IDs.
Scripts under time decision are used to get the exact startting time of the anomaly cases.
util includes all necessary scripts for processing the output from detection to get tracking result using TNT.
All the models used should be placed under model, and you can find the download link for all four models in a txt file under that folder.
You should run the scripts under candidate selection first, and then run TNT to get tracking result. After that please run codes under time decision to get the exact starting time of the anomaly.

Xinyu (Xavier) Yuan
4/9/2019