FURI | Spring 2019
Lowering Image Sensor Reconfiguration Latency on Android
Mobile systems would benefit from the ability to make a tradeoff between image resolution and power consumption when doing vision tasks. While less resolution lowers task accuracy it also saves system power. Currently, resolution changes are a slow process taking over 200ms while also dropping frames. However, researchers propose a system to lower that latency to less than 100ms and without dropping any frames. Researchers aim to seamlessly transition between resolutions with only the already captured frames in the pipeline stopping an immediate change. This would allow mobile vision tasks to minimize power while maintaining acceptable levels of task accuracy.
Alexander C Shearer
Hometown: Chandler, AZ, United States
Graduation date: Spring 2020