About the model weight loading in training stage

[cyang-kth]

In the training script config.model_path is used for loading weights.

https://github.com/jinfagang/keras_frcnn/blob/5a8610c2ef45b2fb38fc2fbebf4c726e680b8b3e/train_frcnn_kitti.py#L91-L94

However, finally the config.model_path is used for export.

https://github.com/jinfagang/keras_frcnn/blob/5a8610c2ef45b2fb38fc2fbebf4c726e680b8b3e/train_frcnn_kitti.py#L240-L243

Can you explain these lines?

I guess that you copy the codes from https://github.com/yhenon/keras-frcnn, where the original code is

try:
	print('loading weights from {}'.format(C.base_net_weights))
	model_rpn.load_weights(C.base_net_weights, by_name=True)
	model_classifier.load_weights(C.base_net_weights, by_name=True)

Can you explain this change? @jinfagang