Future Forward

A. IMPROVED GRAPHICS

Being new to Android development, rendering crisp and informative graphics and statistics was a challenge for all of us. In its current state, Slixstream app is able to render concentric bounding boxes to represent a slipstreamable area behind fast moving objects. We plan to understand the android graphics stack that could help us draw better 3D bounding boxes and alert notifications for the rider.

B. SHARPER COGNITION

Beyond detecting stop signs and fast moving objects, we would want to develop multiple modes in the app that could specifically look for significant landmarks such as break zones, hospitals and first aid, alternate slip routes, bicycle stands etc.

C. FASTER, ACCURATE NETWORKS

Currently, we were able to test and deploy variants of MobileNet and YOLO. The network deployments are far from real time performance and generated a throughput of 2-3 frames per second on a OnePlus 6 handset. We would want to explore or design faster and smaller networks that performs well on a much smaller class space than the existing models and on embedded platforms.

D. BETTER SOFTWARE DESIGN

The current version of Slixstream app aggregates a lot of unnecessary code and android activities which eat up a lot of memory on the phone in runtime. Also, the battery life took a severe hit while these applications were being tested. These two factors, at the moment, make this application unsuitable for practical use but open up a whole new avenue to explore.

Report abuse