Arduino App Lab 0.11 recognizes body poses from a camera image
Arduino App Lab 0.11 adds a tool that recognizes body poses from a camera on the VENTUNO Q small computer. A developer could use a raised arm to select an item on a screen. Recognizing a pose does not assess whether a sports movement is performed correctly.
How does it recognize a pose?
The program first locates points such as shoulders, elbows and wrists in the image. Their positions help it recognize sitting, standing or a raised arm. This is called pose estimation; the new tool processes the images locally on VENTUNO Q.
What is available?
Arduino announced the feature on 22 October. Built-in poses need no separate training step. Developers can also supply photos for a custom pose; the documentation calls for varied examples with the body visible.
What can go wrong?
Classification becomes unreliable when necessary body points are outside the image or cannot be located correctly. Detecting “arm raised” alone does not establish correct exercise technique or a therapeutic result.