Authors: Johannes Bartl 





- Python: 3.8.10
- python packages listed in requirements.txt For training a GPU is recommended - we used:
- Hardware: NVIDIA GeForce RTX 3070 (8GB VRAM)
- CUDA Toolkit: 11.2 (Required for reproducibility)
- NVIDIA Driver: 575.51 (Supports CUDA 11.2+)
- Clone the repository:
git clone https://github.com/fabrylab/AntarcticFurSealDensity.git cd AntarcticFurSealDensity - Install dependencies:
pip install -r requirements.txt
This repository contains the code, models, and analysis scripts associated with the paper "Colony matters: How density shapes predator access in two Antarctic fur seal (Arctocephalus gazella) colonies".
training.ipynb: Pipeline to train the YOLOv8 neural network.analysis.ipynb: Main analysis script for generating statistical results and figures.
model.pt: Trained YOLOv8-large weights file used to predict the full dataset.
FWB.csv: Neural network detection results for Freshwater Beach.SSB.csv: Neural network detection results for Special Study Beach.
FWB.kml/FWB.png/FWB_v1.cdb: Georeferencing and annotation data for Freshwater Beach.SSB.kml/SSB.png/SSB_v1.cdb: Georeferencing and annotation data for Special Study Beach.both_areas.kml/both_areas.png/both_areas_v1.cdb: Combined map data.
The raw imagery and primary annotation databases are hosted on Zenodo.
Dataset DOI: 10.5281/zenodo.18955385
The Zenodo dataset includes:
- Raw Imagery: 110 high-resolution
.jpgimages. - Annotations: Corresponding
.cdb(ClickPoints) databases containing ground-truth annotations.
The analysis.ipynb notebook reproduces the following key results:
- (a) Temporal trends in abundance at Special Study Beach (Figure 2)
- (b) Abundance ratios of birds to pups between colonies (Figure 3)
- (c) Demographic patterns in seal density (Figure 4)
- (d) Spatial associations between birds and pups (Figure 5)
