Total 2030 files were extracted from a video using openCV. The first part of the file contains normal events and the last part contains an amolaous event. For the training purposes only the normal events were considered.
#Process: I used Autoencoder Based system for training the data. I extracted features using a sequential model containing 7 layers of convolutional and maxpooling layer. I used mean square error (MSE) as a loss function. The more detailed aproach is stated in "Steps for anomaly detection.pdf" File