Novelty/Anomaly Detection for Images in Deep Learning

Question

I have only "Positive" class image data from an experiment and the task is to train them using a deep learning network (without having to define features from the image). Anything which is slightly different from the above category should be classified as an anomaly/outlier? [Taking negative class data from the same experiment is quite a difficult task.

I believe that this is entirely not an unsupervised learning because I know the labels of one of the classes. I tried using OneClassSVM for this without defining any features. Also, tried using Convolutional Autoencoders from this "https://github.com/artem-hryb/cnn-anomaly-detection-keras" link and Deep Autoencoders from "https://github.com/otenim/AnomalyDetectionUsingAutoencoder". Some threshold value was set based on training the above models and then it was used to predict if the given image is an anomaly or not. The predictions were not correct.

Could someone please enlighten me on what could be the best solution to try out?

Answer 1

You can use a pretrained CNN models to extract a feature vector for the images and after that train an AutoEncoder on that feature representation. This helps you to easy learn the higher-level differences compared to directly training an AutoEncoder on pixels.

If you want to learn Artificial Intelligence, then check out this Artificial Intelligence Online Course by Intellipaat.