Example with Tensorflow to classify images of emergency vehicles into three classes. Classes:

• fire
• rescue
• police

We've divided the models into 3 project files. Each project files has it's own model.

Findings & Test:

• we are not that good in differentiate betwee fire and rescue cars but good in classify the police cars
• by decreasing the batch size we got worse results (accuracy)

Findings & Tests:

• we decreased the count of filters used
• by decreasing the batch size we got the best results with smaller batch size(16) (accuracy)

Findings & Tests:

• we used a RESNET (many layers)
• by using the loss and accuracy we find out that the model would need a bigger batch size and more training samples