WuJie1010 / CitySCENE2020-Anomaly-Detection

This repository contains source codes of SYSU-BAIDU for CitySCENE Challenge 2020

Modularized Framework with Category-Sensitive Abnormal Filter for City Anomaly Detection (ACMMM Multimedia Grand Challenge 2020)

This repository contains source codes of team SYSU-BAIDU for CitySCENE Challenge 2020, and the technical details please refer to the paper "Modularized Framework with Category-Sensitive Abnormal Filter for City Anomaly Detection".

Our framework obtains a 66.41 MF1 in the test set of the CitySCENE Challenge 2020, which ranks first in the specific anomaly detection task.

Requirements

• Pytorch >=1.1.0

• cuda9

• cudnn7.5

Docker for reproducing our results:

Please download the images in link, then

curl -L -o b53a0eea3f76  https://drive.google.com/file/d/1zLqjvXYkHULTByK5XH-vRDa7jRpq-iSF/view?usp=sharing
docker load b53a0eea3f76 
docker run --gpus all --network none -ti -v <testset_path>:/testset:ro  b53a0eea3f76 /bin/bash
bash /home/example.sh
cltr+d (exit)
docker cp aecd3ec1e4d4:result/task1 ./           
docker cp aecd3ec1e4d4:result/task2 ./

The code is released in the docker.

If you have any questions or issues in using this code, please feel free to contact us ([e-mail])