Sustainably Automate Manual Cell-Staining with Innovative Open-Sour...
summary
AiCSD automates toxic cell staining by replacing wet stains with virtual staining AI-models and running inference on a two-node edge pipeline (OEM + Gateway) using Go services and Python ML code. It packages models with BentoML into Docker microservices, optimizes them with Intel OpenVINO, connects services via EdgeX Foundry, and secures file transfers with an SSH tunnel pattern. The reference runs ~20 containers, processes TIFF microscope images, shipped as open source in June 2023, and deployed in Beckman Couter labs (Denver, Seattle).
AI-written summary · about
chapters (16)
- 0:00Automated cell-staining project intro
- 1:04Speaker intros and agenda
- 2:06IoT and computer vision growth
- 3:13Biopharma AI deployment challenges
- 4:39Cell segmentation use case
- 6:04Project goals and collaboration
- 7:07Solution high-level architecture
- 8:32Architecture breakdown
- 15:38UI and file management
- 17:24Pipeline management with BentoML
- 18:05System demo walkthrough
- 29:25Project challenges and learnings
- 32:34Security in distributed systems
- 34:40Project conclusion and outcomes
- 35:48Future work and extensions
- 36:08Q&A
related talks
36mAI-Powered Real-Time Quality Control
21mCreating Markets, Creating Trustworthiness: Standards for AI
43mGo Microservices for ML at the Edge
11mDemystifying AI in Smart Manufacturing
31mEnabling MEC platforms in Healthcare & Smart Hospitals
44mNext-Gen Therapies
28mEnd to end AI developer journey with containerized assets using Intel DevCatalog, Intel DevCloud
10mRealizing the Power of AI at the Edge