Several open source (BirdNET/Merlin/etc) and commercial projects for bird species identification, coupled with the birding boom of peak COVID, have ushered in a new era in human attention to, and appreciation of, our avian neighbors. However, these systems usually stop at exactly the species level; no aids are available to help us better understand individual visitors. Know Your Corvid aims to use multimodal learning to increase the identification resolution (based on both academic and citizen-originating training data) to reveal the behavioral patterns of specific individuals, encourage even deeper attention, and perhaps eventually build a better modality of understanding between these highly intelligent creatures and the humans whose built environments they now inhabit.

The project is still in very early stages, but I will present the work to date, offer paths of contributing (including a proposed decentralized training approach), and attempt to tackle the thorny question of where this domain of (frankly) surveillance might fit into the privacy-first ethical frameworks that this community champions.

We will try to leave as much space at the end for a discussion if there is interest.