@Pedro_Fernandes The input should be determined by your model, not whatever data you want. This part is no longer within the scope of technical support
@Frank I understand my main question is regarding the conversion which is not really covered in the documentation:
If my model has fp32 as input and fp32 as output when in tflite what kind of data do I have to input to the converted one? int 8 or fp32 and do I have to manually normalize this data or will the conversion take care of this inside the mode? and are the mean values still 128 128 128 128 to make it be between -1 and 1 ?