Problem: scales in test.py

Could you explain how to choose the scales constant in test.py?

The model is loaded in retinaface.py in __init__, (around line 158):

`image_size = (640, 640)`
`self.model = mx.mod.Module(symbol=sym, context=self.ctx, label_names = None) elf.model.bind(data_shapes=[('data', (1, 3, image_size[0], image_size[1]))], for_training=False) self.model.set_params(arg_params, aux_params)
`

So I expect to only being able to provide inputs to the network of shape [1, 3, 640, 640]. Am I wrong?

However... then detect() in retinaface.py takes the scale (which is defined in test.py) as input and does some operations, which lead to an input size that could very well be different than [1, 3 ,640, 640].

How should we choose the constant scales = [..., ...] in test.py?

This question was already asked in issue #684 , with many people supporting the question, but no answer was provided. 请指教 @nttstar @jiankangdeng