A short project to get familiar with the MobileNet neural network model. I tweaked and trained the model to classify images into 10 types of fruit.
I collected and prepared my own training data to make sure that I understand what the training data should look like. After training for only 30 epochs with a training batch of 2000 images and validation batch of 500 images, I got a loss value of 0.0039 and an accuracy of 0.9985, which resulted in a 100% success rate when testing on a batch of 200 images.
I saved the model and training data into a H5 file to use in this program. You can put any images of the 10 types of fruit in the input/images folder. When running the program it will classify all the images and print it to the terminal.