Bodypix & PoseNet Practice
1⃣Considering the Model and Data Biography and reflect
I’m a little bit curious about how these model and data were created. We can see how intelligent and interactive these models are when we’re trying these machines, and they are quite different from the examples we referred to about teachable machines. When we are experimenting with the teachable machine, it’s not hard to see that this artificial intelligence only works on specific occasions, which refers to the appearance of the user and the background (these elements should be almost the same with what the machine are taught at the beginning). But the artificial intelligence in Model and Data Biography seems to have wider applicability like they are easily track the pose of every user or the gesture of the user. How they are trained to have such high autonomy that it’s like the developers have trained it with thousands of different models in disparate occasions. Did these models really go through such a cumbersome process in that case, or was there more technology involved within the progress, this fascinates me a lot.
2⃣My own practice on the Bodypix & PoseNet
Since we all have learnt a little about them on the class and we found that it’s quite difficult to write our own codes, so I plan to practice applying them on my computer, if it works I’ll try further exploration.
I mainly tried Bodypix and PoseNet on my computer and they all work perfectly, so I briefly add some editions to the code to add the PoseNet code part into the Bodypix model’s code, looking forward to draw a colorful figure on the canvas with eyes and nose attached on it (following the motion of the user)
What I find is that they actually can work together but in a high latency, I think it might caused by the model of bodypix since it worked in a high latency separately, and the circles which symbolize the eyes and the nose cannot immediately follow the motion of the user. I really look forward to figuring our how to make that process more efficient and quicker in the latter class.
Practicing video
P5 code link
https://editor.p5js.org/eddiiil/sketches/BIkY55RiH