This is the code for the CVPR'19 paper "Learning Regularity in Skeleton Trajectories for Anomaly Detection in Videos".

Environment Setup

First please create an appropriate environment using conda:

conda env create -f environment.yml

conda activate tbad

Download Data

Due to space constraints in the Github repository, please download the data from the following link and place the data folder in this directory.

Link: trajectories

Test Pre-Trained Models

To evaluate pre-trained models run the evaluate.py script. Some examples:

Evaluate MPED-RNN on all cameras individually and all cameras combined on HR-ShanghaiTech.

python evaluate.py --gpu_ids 0 --gpu_memory 0.2 combined_models ./pretrained/CVPR19/ShanghaiTech/combined_model/_mp_Grobust_Lrobust_Orobust_concatdown_ ./data/HR-ShanghaiTech/testing/trajectories ./data/HR-ShanghaiTech/testing/frame_level_masks --video_resolution 856x480 --overlapping_trajectories

Evaluate MPED-RNN on HR-Avenue.

python evaluate.py --gpu_ids 0 --gpu_memory 0.2 combined_model ./pretrained/CVPR19/Avenue/combined_model/_mp_Grobust_Lrobust_Orobust_concatdown_/01_2018_11_13_06_36_20 ./data/HR-Avenue/testing/trajectories/01 ./data/HR-Avenue/testing/frame_level_masks/01 --video_resolution 640x360 --overlapping_trajectories

Train Models from Scratch

To train a model from scratch you should look up the model's configuration options using the option --help on the training.py script. Here is one example:

Train MPED-RNN on the ShanghaiTech data set.

python train.py --gpu_ids 0 --gpu_memory 0.1 combined_model ./data/HR-ShanghaiTech/training/trajectories/00 --video_resolution 856x480 --message_passing --reconstruct_original_data --multiple_outputs --multiple_outputs_before_concatenation --input_length 12 --rec_length 12 --pred_length 6 --reconstruct_reverse --cell gru --global_hidden 8 --local_hidden 16 --output_activation linear --optimiser adam --learning_rate 0.001 --loss mse --epochs 5 --batch_size 256 --global_normalisation robust --local_normalisation robust --out_normalisation robust