Real Time Detection of Anomalous Activity From Videos (mainly crime actvity). Images of the video is trained using AutoEncoder to get the imtermediate feature representation of image & applied svm model for the bag of such features to detect the anomaly & LSTM to detect the type of Anomaly.

A hybrid deep learning framework for automated diabetic retinopathy detection combining EfficientNetB0 with Swin Transformer attention mechanisms. Features Bayesian uncertainty quantification through Monte Carlo Dropout, explainable AI visualizations with Grad-CAM, and specialized preprocessing techniques.