home projects sketches open source talks

Tensorbot

Tensorbot, nicknamed Tbot, is an autonymous robot that drives around and uses computer vision to speak out what it sees

Tensorbot is a project I built during my residency at sfpc. It is a rasberrypi connected with motors to drive around. For sensing the environment, it uses sonar and a camera. Tbot drives around autonomously searching for object it recognizes. It using deep learning and tensorflow to monitor the camera feed. The machine learning model was trained off of 15 million images of common objects from daily life. Tensorbot was displayed at an art show at sfpc where participants would take printed pictures of common items and would place them in front of tensorbot to recognize. Tbot was buitl in collaboration with Ruby Childs

Next Steps

Tbot was an experiment to see what is possible with machine learning running on small devices. I would like to further develop out tbot into a platform for others to learn and experiment with robotics and machine learning. We envision better speech synthesization and additional features such as facial recognition, motion detection and increased mobility. We may consider making Tbot easy to integrate into any robotic device, such that it can be used as a component for a drone or similar.

Source code can be found here