• Add a Sequence processor.
• Add ability to compose subflows.
• Add processors for:
  • Top-down human pose estimation.
  • Semantic segmentation using tensorflow published models.
  • Pose annotation of images.
  • Mask annotation of images.
  • Basic image manipulation operations.
• Reimplement batch execution engine so that it uses microbadges.

• Logging of events, flow performance and errors.
• Allow user to determine what to do when in presence of errors:
  • Stop flow
  • Drop error frames and continue
  • Other?
• v0.4 is the ground work for v0.5 and v0.9

• Commandline tools to detect bottlenecks by reading logs.
• Commandline tools to debug errors in flow.

• Automatic detection of bottlenecks

• Automatic scaling/decrease of resources in flow without stopping it.
• Will require some changes and augmentations to current execution engines.

• Processors:
  • SLAM
  • Advanced tracking techniques.
  • Others

• Flowboard: Something similar to tensorboard, but for flows.

• Nothing new to add here.

. . .

• Distributed version of Videoflow