Project EyeCommander is a machine learning application to help individuals with a high percentage of paralysis to interact with technology through their eye movements. This is a machine learning project developed for AceCenter, a UK charity that specializes in Augmentative and Alternative Communication and Assistive Technology.
People with a high level of paralysis typically aren’t able to interact with hand/touch-based interfaces and, therefore, struggle with many daily tasks. The main outcome expected for this project is to deliver a MVP system that is capable of using a consumer-grade camera to detect voluntary eye movements (up, down, left, and right) and output a 4-bit signal that can be ported for any application the charity sees fit. For this, we use Docker for containerization, FastAPI for handling module requests, TensorflowLite/Torch depending on the success of Computer Vision experiments, and Python for most BackEnd programming.