yarsmolin

@shashichilappagari, it seems that for 640x640 resolution, EdgeTPU support relies on SLICE2 to bypass the large tensor size restriction, particularly in reshape operations. If I use my custom trained Yolov8...

@khatami-mehrdad Yes! Thanks for an answer. However I need to somehow trained a model for this preprocessing. So far I understood that I need basically implement SpaceToDepth layer ``` import...

@shashichilappagari Thanks for your answer. Btw, I watched your great presentation at ultralytics conference, which spark my interest in DeGirum. So, back to discussion, for example, I trained a YOLOv5m...