Investigating Target Class Influence on Neural Network Compressibility for Energy-Autonomous Avian Monitoring
Nina Brolich, Simon Geis, Maximilian Kasper, Alexander Barnhill, Axel Plinge, Dominik Seuß · 2026
Read Investigating Target Class Influence on Neural Network Compressibility for Energy-Autonomous Avian Monitoring by Nina Brolich, Simon Geis, Maximilian Kasper, Alexander Barnhill, Axel Plinge, Dominik Seuß (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-17751.txt.
Description
Nina Brolich and team study how the number of target classes affects neural network compressibility for energy-autonomous bird monitoring. Biodiversity loss makes wildlife monitoring vital, and birds suit the task since their songs identify them well. Manual counting is costly, while passive acoustic monitoring records soundscapes over long spans for later analysis.
Machine learning speeds that analysis, yet standard models demand heavy compute. The team instead runs models on cheap microcontroller units in the field, under tight hardware and energy limits. They trained and compressed models for varied numbers of target classes to test multi-species detection on edge devices. Results show strong compression rates with minimal performance loss, plus benchmarks across hardware platforms and a feasibility check for energy-autonomous deployment.
Catalog record
| CREATOR | Nina Brolich, Simon Geis, Maximilian Kasper, Alexander Barnhill, Axel Plinge, Dominik Seuß |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 22 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-17751