Overview: For one of our clients in the medical startup sector we have developed a unique prototype of an embedded platform performing computer vision processing based on Convolutional Neural Networks. Object detection together with depth estimation algorithms analyzing a real time video stream on a GPU processing platform provided valuable feedback for the visually impaired

Technologies: Python, OpenCV, TensorFlow, ONNX

Objectives: Develop an AI-based object detection system running on a GPU platform integrated with a depth-reading camera. Video frames passed through segmenters were trained first to recognise surfaces (roads, sidewalks) and then moving objects (cars, people) based on proximity and potential hazard