Deep Learning

We use representations from a teacher to directly train the student network, and prove that it leads to a quicker convergence and better performance.

We use state-of-the-art OCR and NMT techniques to develop an ingenious method of translating books and other documents from one language to another with high-accuracy.

A simple application developed in Flutter that enables transcribing information from images using Google MLKit For Flutter

This is an unofficial implementation of the paper “Temporal Latent Bottleneck - Synthesis of Fast and Slow Processing Mechanisms in Sequence Learning” by Didolkar et. al., developed jointly by Aritra Roy Gosthipaty and Suvaditya Mukherjee