Projects

  • Comparing AI computer vision with Human Labeling

    We have AI results and volunteer contributed human labels for approximately 60,000 images.
    This post explores the relationships between human error and ai errors.

  • Deploying a computer vision model

    Training and publishing a computer vision model on the web.

    Try It Here!

  • Building a Webapp for Species Geofencing

    One of the issues we experienced with speciesnet classification was results where the species was not valid for our location. In speciesnet - specifying a geofence coordinate will cause missclassifications to move up to the highest level hierarchy - in most cases, this classifies an animal as “animal.”

  • Catch A Vibe Code

    I joined an ai for conservation slack group recently. In that group, one of the developers of SpeciesNet (camera trap computer vision) and the creator of the MegaDetector object detection model, Dan Morris, hosted a “Vibe coding party” this week.