dwiniarska/Bangwa-Warbler-Acoustic-Recogniser

Binary spectrogram classifier for Bangwa Warbler vocalisations using a ResNet18 architecture. # Bangwa-Warbler-Acoustic-Recogniser Binary spectrogram classifier for Bangwa Warbler vocalisations using a ResNet18 architecture. This repository contains a PyTorch pipeline for binary classification of spectrogram images using ResNet18. It includes training, validation, model selection based on AUPRC, and a testing pipeline. Although this pipeline was also used for experiments with deeper ResNet architectures (34/50/152), the public version here includes only the ResNet18 implementation. Only minimal code adjustments are needed to swap architectures. ## Publication and dataset This repository is part of the manuscript: **Winiarska, D., Budka, M. _Reproducing a year-round expert survey with deep learning_.** Datasets used with this code are available at Zenodo: Bangwa Warbler Acoustic Rec… --- ## Folder Structure This repository expects the following layout: ``` . data/ train/ positive/ negative/ val/ positive/ negative/ test/ *.png results/ (generated automatically) ``` --- ## Installation Install dependencies: ``` pip install -r requirements.txt ``` --- ## How to Run Run the entire pipeline (training → model selection → testing): ``` python run.py ```