Damaged package detection

The brief

Damaged parcels lead to returns and complaints. Train an image model that tells damaged from intact packages in photos, from more than one angle, so they are caught before dispatch. Good looks like: high recall on damaged packages, a test on unseen packages, and a demo for warehouse staff.

What you will build

1. Explore the intact and damaged package images
2. Fine-tune a pre-trained image model
3. Combine views from two cameras
4. Evaluate recall and false alarms
5. Build a simple check-station demo

Free, open data to build it with

• Industrial quality control of packages ↗ Kaggle

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.