Mini-project
Pharma & Life Sciences
Tablet and capsule defect inspection
Spot chipped, cracked or discoloured tablets and capsules from inspection images.
The brief
Visual inspection of tablets is slow and tiring for people. Train an image model that finds chipped, cracked, contaminated or discoloured pills and capsules, learning mostly from good examples because defects are rare. Good looks like: detection of each defect type, a low miss rate, and a heat map showing where the defect is.
What you will build
- Explore the good and defective images
- Train an anomaly-detection model on good images
- Compare with a supervised classifier
- Evaluate misses and false alarms per defect type
- Show heat maps of the defect location
Free, open data to build it with
Check each dataset's licence before you publish your work. Your mentor gives your team its own version of the brief (data, scope or region), so every team's project is its own.