[2021.1.28] Deploy WG: Strategy & Governance monthly

Thursday, January 28th 2021, 8:00 AM PST - 11:00 AM EST - 5:00 PM European time

Attendees:

1. Karthik Ramadass, 2. Kilian Hett, 3. Jason Crane, 4. David Bericat, 5. Dana Groff, 6. Raghav Mani, 7. B. Selnur Erdal, 8. Bennet Landman, 9. Jorge Cardoso, 10. Yuankai Huo, 11. Brad Erickson, 12. Marco Nolden, 13. Klaus Kades, 14. Ralf Floca, 15. Krishna Juluru, 16. Jayashree Kalpathy-Cramer, 17. Pouria Rouzrokh, 18. Daniel Rubin

Recording:

Action Items (AIs)

New ones

• AI1 - Document conceptual system architecture (Core WG) - [PENDING]
• AI2 - Connect with Evaluation, Reproducibility, Benchmarks WG on System architecture, Input (“trained model standard”), Execution (“Inference artifact standard”) and Output (“inference results standard”) (Lena + MONAI Deploy Core WG) - [PENDING]
• AI3 - Connect with I/O imaging WG on DICOM (Stephen + MONAI Deploy Core WG) - [PENDING]
• AI4 - Connect with Data imaging WG on DICOM (Brad + MONAI Deploy Core WG) - [PENDING]

Pending from last meeting

• AI1 - Publish notes and scope publicly. (DB) - [COMPLETED]
• AI2 - Identify what WGs are stakeholders. Map out dependencies (JC) - [COMPLETED]
• AI3 - Create core WG. (sign up!). Meet 60 mins weekly or 90 mins bi-weekly. - [ALMOST DONE]

AGENDA (+Notes)

SUMMARY:

• Vanderbilt presented their Model Zoo.
• Team gave guidelines for Core WG kick off.

DETAILED:

(15’) Update on pending AIs from last meeting

(20-25’) Vanderbilt U. demo [Bennett + team]

• Model Zoo + existing environment
• Lessons learned + recommendations for open AI deployment API

(20-25’) Core WG

What do we need to make progress

• INPUT: Model trained - format? Data - format (i.e. DICOM) and where is it?
• EXECUTE:
  • What is my executable artifact?
  • Should we define one?
  • How do we package it?
• OUTPUT:
  • Result formats?
  • Consumers? What do they speak? (i.e. DICOM)

Who is:

• Responsible
• Accountable
• Consulted
• Informed

Master document

MONAI Deploy WG - master doc