Wildlife experiments 🦌

This repository contains a variety of experiments implementing the analytic pipeline for wildlife image classification designed in wildlife-ml. Our approach is two-fold, combining improved strategies for object detection and image classification with an active learning system that allows for more efficient training of deep learning models. Please also see https://github.com/slds-lmu/wildlife-ml for a detailed description.

Citation

By using this repo, please cite our paper Automated wildlife image classification: An active learning tool for ecological applications, here is a bibtex entry for it:

@article{bothmann_et_al_2023,
	title = {Automated wildlife image classification: {An} active learning tool for ecological applications},
	volume = {77},
	issn = {1574-9541},
	url = {https://www.sciencedirect.com/science/article/pii/S1574954123002601},
	doi = {10.1016/j.ecoinf.2023.102231},
	journal = {Ecological Informatics},
	author = {Bothmann, Ludwig and Wimmer, Lisa and Charrakh, Omid and Weber, Tobias and Edelhoff, Hendrik and Peters, Wibke and Nguyen, Hien and Benjamin, Caryl and Menzel, Annette},
	year = {2023},
	pages = {102231}
}

Installation

We provide our full Python environment in env.yml.

Workflow

01: Folder structure

├── configs
    └── cfg.son
├── data
    ├── images_megadetector.json
    ├── ...
    └── labels.csv
├── results
└── scripts
    ├── execute_experiments.py
    ├── ...
    └── run_experiments.sh

02: Order of execution

• The scripts are executed in the following order:
  • prep_config.py
  • prep_data.py
  • execute_tuning.py (optional)
  • execute_experiments.py
• Each Python script has an accompanying bash file that allows for convenient, step-by-step command-line execution with minimal user input.
• prep_config.py specifies user preferences, in particular:
  • img_dir: path to directory where images are stored. Please note: all images should be in a single directory, allocation to different datasets is done via separate files containing the respectively relevant keys.
  • data_dir: path to directory where intermediate data-related files shall be stored (directory must exist).
  • result_dir: path to directory where results shall be stored (directory must exist).
  • active_dir: path to directory where data related to active learning (e.g., images to label in current iteration) shall be stored.
• Further settings in prep_config.py may be adapted if desired.

Coming soon

We are currently working on deploying a server to host data and the active-learning system.