yashjakhotiya / Anomaly-Detection

UCSD Anomaly Detection Challenge with CNN Autoencoders and 2-layered LSTMs

Anomaly-Detection

This project is trained on the UCSD dataset. The code has been kept in modular format, with the following modules -

1. data_loader.py - Contains two tf.keras.utils.Sequence classes - CNN_train_data_loader and CNN_test_data_loader to load batches of images and labels
2. paths.py and hyperparams.py - Contains paths and hyperparams needed for the model
3. logger.py - Logger callback
4. timer.py- Timer callback
5. train.py - Entry point for the code; starts the training sequence
6. model.py - Contains two models - a CNN Autoencoder and a 2-layered LSTM

CNN Autoencoder

A CNN Autoencoder is trained to learn latent space representation of frame images

Stacked LSTM

A 2-layered LSTM is trained to predict if the sequence of frames contains an anomaly. Latent space representation of each frame is passed to the stacked LSTM.