On-device ML for Object and Activity Detection
Established: January 8, 2015
- Delivered full feature-extraction IP (Verilog and C#) to the Windows Phone 8 product team for integration and shipping with the ISP
- Built (and helped build) end-to-end working prototype systems to demonstrate the benefits and challenges of ambient-aware computing
- Pose detection for law enforcement with body-worn cameras
- Low-power FPGA-based on-device image classification
- LoRa-based networks to transmit data from vehicles over long distances
- Fast XNOR nets with Intel and ARM CPUs for image classification on the edge
- Contributed presentations to the senior leadership team at the Faculty Summit, Build Conference and internal review meetings.
- Disseminated research results through conference papers at DATE 15, DAC 15, MCVMC 15, ISLPED 15 and TMSCS 16
- Filed US and international patents that protect innovations in mapping ML algorithms to edge devices for ambient-aware computing
People
Matthai Philipose
Senior Principal Researcher
Bodhi Priyantha
Researcher