Development of Tools for Data Assessment, Active Learning and Believability for Visual Data
This project should aim at developing the following
1. Data Assessment / Feature Selection tool:
a. To effectively select features (from a given set) that are optimal for machine learning application using supervised (with classifier deep neural network in loop) and unsupervised techniques.
2. AI Validation and Optimization tool:
a. Believability: To quantify the difference in the behaviour of classifier deep neural networks with change of data from the training dataset (on which the algorithm is calibrated) to the actual dataset.
b. Active Learning: To develop active learning techniques to select subset of data (from large amount of unlabelled data), which will maximize the learning of classifier deep neural network models.
3. Application software to utilize the above
a. Design and development of web based application software.
b. Faster access to high dimensional feature vector.
c. Scalable storage and faster retrieval for processing.
Project proposal is open only for startups.