Object Detection and Human Posture Recognition with Pepper Robot

For object detection, I retrained yolov3 object detector on images of some customized items in which I manually gathered, performed image processing, and annotated. I then used Pepper robot sensors and API in order to integrate the trained network with Pepper (video 1).

For human posture recognition, I built a machine learning classifier on top of state of the art human body key points estimator (OpenPose) to classify different human postures, and then integrated it with Pepper robot (video 2).

This work was done as part of my internship at Grenoble Informatics Lab.