Release v1.0

The weights for the pre-trained backbone have been attached to this release. You may download them and transfer the same to your project.

Note

The architecture of the backbone has to be re-built in your project before you transfer.
You may use these weights as a segment in your model as well by setting strict=False while loading the weights to your model.
Since Torch uses dictionaries to store weights, use the same naming strategies for sub-modules. For instance if it is self.linear use the same name instead of self.lin