Pass argument pretrained_dir in build_refinenet to frontend_builder

[msseibel]

Please specify the following information when submitting an issue:

• What are your command line arguments?: None
• Have you written any custom code?: No
• What have you done to try and solve this issue?: Nothing yet
• TensorFlow version?: Not relevant

Describe the problem

Something between bug and feature: See the linked line below: https://github.com/GeorgeSeif/Semantic-Segmentation-Suite/blob/4b7eb0c1a1622367a4378cfd42d3b5b621bcb3e8/models/RefineNet.py#L158 The argument pretrained_dir from the method build_refinenet is not passed towards the method frontend_builder.build_frontend and therefore currently useless.

The argument pretrained_dir is also not passed in https://github.com/GeorgeSeif/Semantic-Segmentation-Suite/blob/4b7eb0c1a1622367a4378cfd42d3b5b621bcb3e8/models/DeepLabV3_plus.py#L75

Are there any reasons for not passing it?

My suggestion: logits, end_points, frontend_scope, init_fn = frontend_builder.build_frontend(inputs, frontend, is_training=is_training,pretrained_dir=pretrained_dir)

[GeorgeSeif]

Great catch! I totally missed that so thanks :)