I would like to take MobileNet and train it to classify different features (deleting any previous training).
I have input the data and formated it to what I think is required. I have also modified the NN to manage the parameters.
I have attached the code rather than pasting it. To me the training data format looks ok but it tells me it's not. Any help or advice would be really appreciated. This is my first NN so I'm not sure what I'm doing wrong.
Just to clarify the functionality of the net, I have deleted the standard labels (softmax layer) and replaced it with 27 binary outputs that switch on or off depending on the image content.