memAE

This is an unofficial implementation of paper "Memorizing Normality to Detect Anomaly: Memory-augmented Deep Autoencoder (MemAE) for Unsupervised Anomaly Detection".

Majority of the code are based on the original repo https://github.com/donggong1/memae-anomaly-detection

Dataset	Paper	This implementation
UCSDped2	94.1	94.0
Avenue	83.3	81.0

Requirements

• PyTorch == 1.4.0

• Python==3.7.6

• ./requirement.sh

Prepare dataset

prepare_data.sh dataset datapath
dataset: Avenue, UCSDped2
datapath: the path that you want to save the data, i.e., /project/anomaly_data/

Train the model

./run.sh dataset datapath expdir
dataset: Avenue, UCSDped2 
datapath: the path that you want to save the data, i.e., /project/anomaly_data/
expdir: the path that you want to save the checkpoint

Evaluate the model

./eval.sh dataset, datapath, version, ckpt, expdir
dataset: Avenue, UCSDped2
datapath: the path that you saved the data
version: experiment version
ckpt: the checkpoint step
expdir: the path that you saved the model checkpoint

If your Avenue dataset is saved under /project/anomaly_data/Avenue/frames/testing/...., run ./eval.sh Avenue /project/anomaly_data/ 0 40 ckpt/ to get the reported performance