2021-ImageRecognition-CMIAEConference

« back to Publications

Greenberg, S. (2021)
Image Recognition for Camera Traps: Making it Work for You. In Scaling Up Camera Trap Surveys to Inform Regional Wildlife Conservation Conference, Columbia Mountains Institute of Applied Ecology, May 18-20. Video, duration 19:25.
Type: Conference presentation (abstract and video).
Note: This is a re-recording of that presentation. PDF is the talk’s abstract.
See also: page 7 of the full proceedings. Presentation base on this August 2020 report.

Download as: Video and Pdf

Conference: Conference page

Abstract.
You may have heard about applying automated image recognition to wildlife camera trap images. The basic idea is that an image recognizer will automatically analyze your images to locate and classify the wildlife species captured within each image. When recognition succeeds, it correctly detects and identifies whether animals are in an image, along with a classification of each detected animal. Various academic papers report what appears to be excellent recognition performance, some in the 90%+ correctness range in detecting wildlife and identifying the correct species. This may make you believe that you should now apply an image recognition system to automatically classify the (possibly) millions of camera trap images you have collected, as the time and cost savings can be enormous. Yet before you do, you should be aware of the nuances of applying recognition to your images. This includes understanding why recognition on your own images may be less than the reported performance, knowing where recognition will likely succeed or fail, and how to adjust your workflow to include a `human in the loop’ to make best use of recognition predictions. The presentation discusses these points, where it illustrates a workflow using the freely available Timelapse image analyser for camera traps (https://saul.cpsc.ucalgary.ca/timelapse/), which in turn incorporates recognition results produced by the Microsoft Megadetector https://github.com/microsoft/CameraTraps/blob/master/collaborations.md

Scroll to Top