Skip to content

Transformations Meeting 2

mjorgecardoso edited this page May 26, 2020 · 2 revisions

Requirements

  • 3D - Needed
  • 4D - Likely (To be discussed)
  • GPU - Needed
  • CPU - Needed
  • Differentiable - Needed when available
  • Parameterisable - Needed
  • Invertible - Needed when available
  • Consistent UX - Beneficial
  • Apply to different pairs - Needed
  • Lazy spatial transformations - Beneficial
  • Take image geometry into account - Needed
  • DAG/Conditionals/Composition - Likely (To be discussed)
  • Caching Intermediaries - Likely (To be discussed)
  • Reproduciblility for randomised transformations
  • API for integration of external packages with working examples

Questions to be discussed next meeting

Dicts vs Lists? DAG vs Compose? Torch Tensor vs numpy?

Clone this wiki locally