• A three day talk and lecture series on applications of Deep Learning on Computer Vision, Natural Language Processing and AI for Social Good & Human Computer Interaction.

Adviosr: Dr. Rahul Panicker

• Developed a Multi-View pipeline for the Visual Weighing Machine (VWM) project for reducing thickness ambiguity from predictions.
• Formulated and implemented a novel Depth-Aware weak-perspecitive camera for Geometry based Deep Learning pipelines.
• Formulated and implemented a model architecture fusing Multi-View with Geometry which can make scale aware predictions for a subject using a supplementary object.
• The overall algorithm increased the performance of the VWM pipeline by over 26% on validation and over 14% on test (MAE).

Adviosr: Dr. Rahul Panicker

• Worked on the Visual Weighing Machine project, used for measuring body mass of new-born infants.
• Used novel latent space representations for human mesh manipulations.
• Created a robust pipeline for acquiring anthropomorphic measurements of human babies from 3D meshes.
• Increased performance of the algorithm by 10% using state of the art Domain Adaptation techniques.

Adviosr: Prof. Debashish Chakravarty

• Developed and implemented various state of the art motion planning algorithms for autonomous driving.
• Integrated various high precision sensors (LIDAR’s, IMU, GPS, etc), with different industrial-grade robot prototypes(Husky and Jackal) and mid-sized vehicles (Mahindra E2O) for developing and deploying several state-estimation and planning algorithms.
• Surveyed and Implemented various Safe Reinforcement Learning algorithms for autonomous control and navigation.
• Developed and maintained the full simulation stack for the autonomous driving system.