ok so that's pretty hard basically i think if i engage object tracking at all it will be helpful to me 1347
hmm so it can be funny to think about but really ideally there'd be a model or encoding of an object that can be used to identify it from video of it .... one would imagine such a model or encoding would also be useful for drawing the object. now, sometimes an object is partially obscured. sometimes its partially obscured (often even) in different parts throughout a video, so no individual frame captures it wholly but -- 1350 in machine learning people turn object pictures into nerfs. and they're pretty heavy for embedded systems. but there's also object identification. and, i thinnk that involves training the identification model on novel objects. maybe that's the way to go here. maybe not uncertain. it seems like really useful ways sometimes can provide for multiple approaches. especially, if those are combineable, that can be very helpful. 1351 1355 I'm looking at this a little and seeing that there's a lot of research in the parts of it, but research around the kind of synthesized whole is not bumping into me as prominently. I'm looking at YOLO a smidge which looks nice. But I'm thinking that this might be something where would want to leave an avenue for there being new research -- either by me or found and integrated -- to make the parts synthesize better together for the purpose. y--1356 1359 ok after looking at yolo a little ... i'm remembering that it's way cooler to track all the user's objects than to know what they are. so we might as well focus on object tracking rather than identification. if needed, maybe we can imagine a hundred robots that all run around staring at objects to know that they're there. alternatively, we could simply say we're deciding to prioritise the handling of in-frame objects first before the handling of those that go out of frame. i guess usually one correlates with a different viewpoint. ok back to normal object tracking ! 1401 1404 so basically opencv can possibly do very good object tracking now: https://learnopencv.com/object-tracking-using-opencv-cpp-python/#opencv-tracking-api and it has a relatively simple high-level interface. the newer your opencv version the better the technology. !!!!! it's fun i got somewhere with the idea T_T
