What's Next for HSF?

Here are some ideas for what is yet to be included in the next major version of HSF:

• HSF
• Complete the documentation
• Remove pytorch and torch.io deps (use DALI?)
• Provide a way to QC segmentations
  • Voxel-wise aleatoric uncertainty.
• Models
• Single shouldn't be part of the bag because it hasn't seen all the data
• Increase the amount of training data
• Enhance data augmentation
• Float16-int8 / Quantization (delayed)
• Quantization Aware Training (delayed)
• Pruning (WIP)

Please note that this list 1/ is not exhaustive, 2/ is not a guarantee that the mentioned features will be included in the next major version.