No, on vim3 you can run the converted thensorflow graph and on NPU, say for example mobilenet ssd v2, at 70FPS. On vim1 you can only run the tflite against the cpu at probably around 6-7 fps.
While tensorflow does not support directly other hardware apart from cuda enabled, there are means to convert the graphs for various inference devices as AML NPU/VIM3
2 Likes
@ufneeme If you use VIM1 . You will lose NPU, If you can’t run with NPU, it’s extremely inefficient .