Structuring automated visual inspections with Django and gRPC
summary
An MLOps prototype uses Django admin + models to define Product → InspectionView → InspectionTask (x,y,width,height) plus Label records, validates rectangles inside the reference image, and stores captured samples in SQLite; a generated gRPC store_image proto accepts view_id and image bytes and returns the new image id, and the demo actually sends images and they appear in the admin. Packaging uses Poetry for fast dependency locking. Emphasis is on traceable data, human-in-loop labeling, and preferring anomaly detection for low-volume SKUs.
AI-written summary · about
chapters (18)
- 0:00Introduction
- 2:07Visual inspection requirements
- 5:58MLOps for robust ML
- 8:25ProDrive AI framework
- 9:50Django features
- 11:16Data models for inspection
- 15:04Django and gRPC setup
- 17:48Django data models
- 20:15Customizing Django admin
- 21:18Migrations and admin user
- 23:00Image labeling and tasks
- 23:43gRPC service with Protobuf
- 25:29Generating gRPC code
- 26:10Running servers
- 26:53Sending images via gRPC
- 27:36Future considerations
- 28:19Key takeaways
- 29:24Q&A
related talks
21mNeedle in the Haystack: Applied ML for Inspections of Power Infrastructure
26mBuilding and Scaling an AI Startup with Async Django
50mBuilding an MLOps strategy from the ground up
45mDjango for AI: Deploying Machine Learning Models with Django with Will Vincent
34mFrom "Vibes" to Production: Eliminating RAG Data Silos with MCP
26mWhat it's really like building RESTful APIs with Django
26mBuilding Data and Machine Learning Applications using Gradio
27mPractical MLOps for better models