A former Intel engineer with expertise in AI, machine learning, and hardware architecture, Phillip Ha is on a mission to bring robots to the masses. He’s doing just that through his venture Robot Wonderland, and by creating practical and useful demos to create awareness of the potential of personal robots.
Moreover, he’s beginning to turn his projects into teaching tools for the next generation of engineers and developers.
His most recent—and timely—project is a mask-detecting robot based on Misty Robotics' open robot platform called Misty. The company was launched by Ian Bernstein, the creative mind behind the educational toy Sphero.
Developers can build skills for Misty that run directly on the platform, or create applications that run on an external device using Misty’s REST API. Sharing and collaboration is encouraged through a community forum.
Ha's mask-detecting robot project took about a week to complete.
Misty comes with a robust set of generic capabilities, including motion control, audio recording and video streaming.
Ha augmented these with his own custom AI/deep learning model using TensorFlow and Keras frameworks that runs on an external device through the RESTP API. “The custom AI/DL models were trained for Misty to use to recognize not only my face, but any person’s face both with and without a mask,” said Ha.
The built-in video camera on Misty is arranged in a portrait format that can recognize any face between a distance of 2 and 6 ft and can detect a face even if it is turned 45 degrees from center. “It makes a determination on whether it sees just the eyes or the entire face, and if it is the latter will speak a friendly reminder,” said Ha.
That’s no small feat. Even Misty’s creators admit that It’s hard to do facial recognition. But Misty comes equipped with a Qualcomm Snapdragon Neural Processing Engine for AI, which can be used to train the robot on objects and recognize them.
From start to finish, Ha estimated that it took about a week to do the planning, gather training data, conduct the training and test the new models, program and test the skill with the Misty robot, and make a demo video to post. He credits the ability to use some of the core program libraries for controlling Misty and his own background knowledge for the quick timetable.
Ha is working on the project documentation now, and once people are out and about again, he plans to take it to events where students have a chance to see it up close. There’s no reason it could not be set up in a public space right away if there is interest.
“Some grocery, retail stores, or restaurants may still want to remind people to wear a mask,” he said. “And in some cases that might lead to a bit of a challenge. But with a cute robot, it may be a different story!”