Technical Reports

For the curious, the technical, and the data-driven, this page is your direct window into our research methods. Here we share our internal "white papers" and technical reports—the same documents our science team uses to build and validate our tools.

You can explore a deep dive into how we built the "BatNoBat Sorter" from the ground up, or read about our specific data analysis protocols. These documents are technical, but we believe in open science and making our work available to anyone who wants to explore the "how" behind our discoveries.

Concordia Bat Project Technical Report: The Bat/No Bat Sorter Program

This report details the creation of our most critical tool: the BatNoBat Sorter. To solve the impossible problem of finding bat calls hidden in terabytes of nightly noise, we developed our own machine learning system from the ground up. This paper describes our "bootstrap" method—how we started with simple nature-based rules, then used our experts to manually verify over 52,000 sound snippets. This massive, hand-curated dataset was used to train a powerful convolutional neural network that now identifies bat activity with near-perfect, human-level accuracy.

Click here to read the full technical report on how we built the BatNoBat Sorter.