Labels like Bounding Boxes need to be scaled according to changes in data dimensions in order to make it conformal to feature space, this is required in feature-level analysis around the target region, some other NN architectures require this to shard feature space for further processing.
Implementation:
There is no need to amend core API, instead a simple extension of related classes in a separate module extension within the moduleZoo is sufficient.
Module Priority:
- Convoution Blocks
- Residual Blocks
- 2D Attention Blocks Possibly not required
Labels like Bounding Boxes need to be scaled according to changes in data dimensions in order to make it conformal to feature space, this is required in feature-level analysis around the target region, some other NN architectures require this to shard feature space for further processing.
Implementation:
There is no need to amend core API, instead a simple extension of related classes in a separate module
extensionwithin themoduleZoois sufficient.Module Priority: