Candy Detection & Counting
Counted once each piece tracked across frames, even when it overlaps.
- Client
- Confidential
- Year
- 2025
- Category
- Computer Vision
- Role
- Design + Build
- Timeline
- 3–4 weeks
The problem
Counting product on a fast conveyor by eye or by weight is approximate. Overlapping pieces and motion blur break naive frame-by-frame counting.
What we built
A YOLO detector plus a tracker that assigns each piece a persistent ID and follows it across the belt, so a piece that overlaps or is briefly occluded is still counted once. Trained on annotated line footage for the specific product and lighting.
The result
An exact, continuous count from one camera over the belt. No scale, no manual tally.
- persistent track ID across frames
- per piecepersistent track ID across frames
- counts at line speed on the conveyor feed
- real-timecounts at line speed on the conveyor feed
- handles overlap and brief occlusion
- count oncehandles overlap and brief occlusion
Built with
- Python
- YOLO
- OpenCV
- Object tracking
- Data annotation
We deliver what we commit.
Tell us what you're trying to build.
We'll come back within 24 hours with honest feedback on scope, timeline and cost, whether or not we turn out to be the right fit.