Machine learning for computer vision in pharmaceutical manufacturing
summary
Claim of an industry-first automated quality assurance system that counts about 370 cartridges per box in real time using a robotic arm, top-down camera rig, on-device ML inference and an Amazon QuickSight dashboard. Prototype arrives from a six-week engagement that uses Amazon S3, SageMaker Ground Truth, SageMaker Pipelines, SageMaker Edge Manager, AWS IoT Greengrass, Timestream and QuickSight. Demo runs at a summit with no extra lighting, proves robustness, and the same pipeline repurposes to agar-plate anomaly detection with autoencoder models and full dataset and model traceability.
AI-written summary · about
chapters (9)
- 0:00Introduction and background
- 4:49Computer vision initiatives
- 9:11Challenges with scaling POCs
- 12:04Pipeline development process
- 18:24Live demo of solution
- 20:33Monitoring and traceability features
- 24:35AWS architecture overview
- 24:58Repurposing pipeline for new use case
- 28:41Future plans and next steps
related talks
29mAmazon Lookout for Vision
01h 00mOpening Keynote
19mDrive transformation through machine learning
51mProfundizando en Aprendizaje Automático para Desarrolladores
36mAI-Powered Real-Time Quality Control
32mAutomating Production Production Level Machine Learning on AWS
41mPractical Aspects of Applying Modern Computer Vision Technologies
14mComputer vision on edge devices