I tried to load the models from https://github.com/CMU-Perceptual-Com... but the example code fails at assert(len(BODY_PARTS) == out.shape[1])
Am I using the right models? In the example a simplified model is mentioned. Does this mean that I need to load this simplified model? If this is the case, where can I find it?
@graspergazon, @berak, usually we try to introduce a new model not only in samples but in tests too. In example, https://github.com/opencv/opencv/blob... . You may find corresponding .prototxts at https://github.com/opencv/opencv_extr... and .caffemodellinks at https://github.com/opencv/opencv_extr... .
Thanks for the data. The code works perfectly fine on ubuntu with opencv 3.4.1. I am trying to run the code on android platform using opencv 3.4.1 but I am getting an error. I have described the error in depth on the Github page. Here is the link:
https://github.com/opencv/opencv/issu...
It would be great if anyone can help me. I can share the android code. Your help will be highly appreciated.
Asked: 2018-02-16 05:07:04 -0600
Seen: 1,108 times
Last updated: Feb 16 '18
hehe, you're exactly the person, i've been looking for, playing with the very same thing !
are you using the COCO , or the MPI model ?
(in the case of COCO, it seems, you have to use
out = net.forward("Mconv7_stage6_L2")(that is, stop before the final Concat layer) , else you get 57 heatmaps instead of the expected 19we'll have to find out, what there is in the other 38 heatmaps, and how to make use of those ;)
i'm somehow suspecting, that the additional 38 heatmaps are the hand parts , but i'm not sure.