Sep 2019 – Dec 2019 · 3 mos · Amherst, Massachusetts

• Identifing Player Action in Tennis using RGB Video Data
• ◦ Created a deep NN classifier to classify 12 types of different strokes using in tennis, trained on THETIS dataset.
• ◦ Used Transfer Learning, ResNetV2 architecture to extract features from frames of video and analysed the temporal dependency in frames using a 3 layer LSTM network.

May 2018 – Jul 2018 · 2 mos · Bengaluru Area, India

• ◦ Developed a prototype setup for automatic calibration of single-wheel projector which was earlier done manually.
• ◦ Designed algorithm requiring less than 15 seconds to capture and process photo sensor data.
• ◦ Removed noise and pulsation effect of the lamp source using cross correlation and detected color transitions using standard deviation map of signal data.
• ◦ Implemented protocols for ADC board initialization and capturing by making API modules