I have trained a model using googlelenet to traning model classification plates of cars and motorcycles. When I run the model on digits, it returns the plate of the car, but when I run code with C++ open image with opencv( version 3.4.7) it’s return a plate of Motorbike. If the problem is the reading of the image input? Beause I read that the Digits using python’s input function is read a image imput, whether that matters?
It’s recommended to check if the preprocessing steps between DIGITs and jetson_inference are the same first.
Ex. mean-subtraction, color format, input value range.