Visual defect checker for cast parts
Inspired by publicly reported work at Bosch
Spot defective cast parts from photos so inspectors only look at the doubtful ones.
The brief
Bosch has written about using AI for optical inspection in its plants, flagging doubtful parts so human inspectors look only at those. Train an image model that separates good from defective cast parts, choose a threshold that sends uncertain parts to a person, and show what share of inspection work that saves. Good looks like: recall on defects near 100% at the chosen threshold, a confusion matrix, heatmaps showing where the model looks, and a simple inspection screen.
Read the case first
What Bosch has said in public about this kind of work, on its own pages:
- How Bosch uses AI in manufacturing (Bosch Media Service) ↗
- Toward zero-defect production with Bosch artificial intelligence (Bosch Media Service) ↗
- AI-based solution for optical defect detection (Bosch Media Service) ↗
Make it your own
Inspect something from a workshop near you: PCB solder joints, stitched garments or packed tablets. Photograph 200 samples yourself (with permission) and test how well a model trained on the open data transfers.
AITeky has no link with Bosch, which has not reviewed or endorsed this brief. The brief only draws on what the company has published, uses open data rather than its data, and is a learning exercise: don't present your project as Bosch's work or system.
What you will build
- Explore the images and balance the classes
- Fine-tune a small CNN
- Pick a threshold with a "send to inspector" band
- Show heatmaps (Grad-CAM) on defects
- Estimate inspector time saved
- Build an inspection screen demo
Free, open data to build it with
Sign up free to get the data and links
It takes a minute with your email. AiTeky Pass (₹999/month or ₹9,999/year) is free for your first 30 days.