Teach Your GO1 to Detect a Treat Using Computer Vision!

The DroneBlocks team is always mining for new tricks to teach our Robot Dog. And any time fresh inspiration comes from a fellow tinkerer – we invite them to join the Curriculum team to share their knowledge with our eager community. That’s just what happened with this course – though in this case…”tinkerer” may be an understatement! Gabriel Margolis is a PHD Graduate Researcher at the MIT Robotics Lab who specializes in quadruped robotics. Gabriel has extensive experience experimenting and innovating with the Unitree GO1. In one of Gabe’s recent studies, he challenged himself to teach the GO1 to dribble a soccer ball! And that’s just part of Gabe’s sweeping resume of pairing Quadruped software and engineering mechanics with mobility and computer learning. We know our DroneBlocks students are still charting their journey to MIT…so we asked Gabe to pare down a specific aspect of his research into a digestible course for our students and educators, and what he put together is truly amazing.

The course is titled Using Computer Vision for Object and Face Detection with Go1. In this course, students will learn “stack overflow” or how to work with Gabe’s provided source code. They will load it onto the Unitree GO1 and edit it to fit their needs, all the while Gabe reviews the how and why. In the first project, students will use the GO1’s onboard cameras to detect a colored object or “a treat.” In the examples, a Red party cup is used, but one can imagine how this skill could be a building block to a robot detecting a soccer ball. Students will learn the power and implementation of object recognition, by programming the GO1 to “bark” when detecting the object. Students will be able to draw comparisons to the everyday technologies that utilize it, as well as conjure future creative uses for the technology. The lessons then go one step further and delve into programming the GO1 to use facial recognition. A fun time lies ahead in programming the GO1 to react when it recognizes a face and again teaching it to “bark” or flash its LED lights. The great thing about this course is that it opens the floodgates to a whole new world of exploration with not just the GO1 but any tech that uses a camera. Not to mention, the first few lessons of this course are accessible to experience even if you do not have a Unitree GO1, as pre-recorded videos are included as downloads in the course and can be used to simulate the exercise.

Throughout the course, you’ll gain hands-on experience in harnessing computer vision algorithms to empower the robot dog’s perception and interaction skills.

We invite all our educators and students to take advantage of this amazing opportunity to work alongside a robotics researcher as passionate and skilled as Gabe. The only prerequisite recommended is a basic understanding of Python and a willingness to learn. This course is brought to you through a combined effort of all our curriculum specialists…from New Zealand to Boston, from New York to Texas. We can’t wait to see what tricks your robot dogs are capable of!

Access the course