Utilizing Deep Active Learning for optimizing SLO (Scanning Laser Ophthalmoscopy) retinal data selection using image embedding vectors

This project is mainly focusing on use Deep Active Learning models to optimize the data selection rules in terms of adding training data and model retraining. DNN are used to detect features on the retinal images thus extract the learned image embedding vector as well as the video sequence embedding vectors. These embedding vectors are then used to design an active learning data query system thus explore the values of different training data.