Create model architecture in model.py file

[alikureishy]

This file will expose a method:

def create_model_for_training(.....):

... and return the model architecture as below:

  return input, output, loss_function, optimizer ...

This will be used by the tl_trainer.py node. Having this as a separate file allows the architecture to be implemented independently from the trainer.

[bessszilard]

I suggest pretrained SSD_GRAPH or FASTER_RCNN. In this case, we have only to fine-tune the existing model.

[bessszilard]

Without any fine-tuning I was able to detect traffic light contours on the image in most cases.

[bessszilard]

The most elegant method is to fine-tune the model, and the CNN will classifier the traffic lights's color too.