DQ Scan a Leaf
- Oct 7, 2025
- 1 min read
Updated: Aug 1
For DQ Canada's Scan a Leaf Snapchat Lens, I built and trained a custom computer vision model capable of recognizing 10 different maple leaf species. The model was trained in Python using a custom training pipeline I developed in Jupyter Notebook, leveraging 3,000+ manually labeled images and GPU acceleration on my home server before being integrated into Lens Studio for a real-world AR campaign.
Campaign Results
2.5M+ Snapchatters engaged with the Lens
10+ seconds average Lens playtime (above Snap Canada AR benchmarks)
54% higher reach with Snap First Commercial vs. auction-based commercials
4%+ click-through rate, outperforming category benchmarks
Snapchat: https://www.snapchat.com/unlock/?type=SNAPCODE&uuid=78c845c00d774baaa68a538f577b5b1e&metadata=01
Client: Dairy Queen
Creators: Benjamin Paruzynski, Bex Bartels
Management: Subtropic
