I'm currently developing a SLAM algorithm that I would like to run on K210 and for this, I need to input 2-3-more images to the network at the same time (stacked by channels).
I've browsed through the code and got as far as lib_sipeed.a. I can see it could be possible to modify py_image to be able to stack images(lots of workarounds) or modify maixpy_kpu to accept several images(fewer workarounds), however, I'm not sure how RGB565 is being split into individual bytes inside the closed source library which is a blocker to my investigations...
I'm currently developing a SLAM algorithm that I would like to run on K210 and for this, I need to input 2-3-more images to the network at the same time (stacked by channels). I've browsed through the code and got as far as
lib_sipeed.a
. I can see it could be possible to modifypy_image
to be able to stack images(lots of workarounds) or modifymaixpy_kpu
to accept several images(fewer workarounds), however, I'm not sure how RGB565 is being split into individual bytes inside the closed source library which is a blocker to my investigations...Any suggestions?