Snapshot Wisconsin: Oh Deer – Predicting Population Dynamics from Trail Cameras

Projects
ML Marathon
MLM26
Computer vision
Camera trap
Ecology
Image classification
Deep learning
Author

Ryan Bemowski

Published

August 15, 2026

Snapshot Wisconsin: Oh Deer is a challenge in the 2026 Machine Learning Marathon (MLM26), hosted by Ryan Bemowski. Snapshot Wisconsin has collected over one hundred million motion-triggered trail camera photos from across the state, with volunteers worldwide classifying species and population dynamics tags – but much of the information in these images has never been converted to a structured format. This challenge asks participants to build tools that predict population dynamics tags from deer images, supporting ecological research and wildlife management.

Challenge design

  • Task: Predict two population dynamics tags for each deer image – Antlers (antlered or antlerless) and Age (adult or young). For extra credit, extract additional useful information: time of year or day, whether animals are alert or relaxed, land cover characteristics, or anything else valuable to ecologists.
  • Data: 4,880 labeled trail camera images of deer plus a ground-truth CSV including bounding box locations for each deer – provided the way you’d receive data in a real-world project: filtered and cleaned, but not perfectly organized for any particular approach.
  • Evaluation: 150 points total – F1 score on the Antlers and Age predictions (100 points) plus a usefulness rubric (50 points) covering additional predictions, inference speed, compute efficiency, and ease of use for research software engineers.

Comments