inputs:
  PIL_mode: L
  ext: jpg
  type: images
targets:
  type: integer
type: vision