🌿Species Identification and Coverage Estimation from Ground-level Imagery for Vegetation Monitoring 📷

This repository contains the data and code used in Müller, Puliti, and Breidenbach (2025) to train and apply deep learning models for species coverage estimation using ground-level imagery.

It includes:
✅ A YOLOv8 object detection model for detecting frames in images and one species instance segmentation model for species identification and identifying and segmenting species.

✅ Method to parse instance segmentation masks to species-specific coverage estimates in images.

✅ The data used to train and evaluate the models

🚀 Workflow Overview

This demo provides a step-by-step approach for training and applying the models:

1️⃣ Training:

• Train two models using labeled images:
  • Frame Object Detection (dataset: Frame_data)
  • Species Instance Segmentation (dataset: Species_segmentation_data)

2️⃣ Confidence Optimization:

• Optimize the confidence threshold based on downstream cover estimation performance.

3️⃣ Inference:

• Predict on test images (Species_cover_data_test).

4️⃣ Evaluation:

• Compare predictions with field estimates (Field_data_NFI).

📌 The code has been tested on Windows with Python 3.10.

🛠 How to Run the Demo

Follow these steps to set up and run demo.ipynb:

📖 How to Cite

If you use this work, please cite:

Müller, P., Puliti, S., & Breidenbach, J. (2025). Towards Enhancing Field-Based Vegetation Monitoring: A Deep Learning Approach for Species Coverage Estimation from Ground-Level Imagery. Methods in Ecology and Evolution.

📜 License

This project is licensed under the GNU Affero General Public License v3.0 or later (AGPL-3.0-or-later).

🔹 Key points of this license:

• You are free to use, modify, and distribute the software.
• If you modify and deploy this software (even as a web service), you must share your modifications under the same AGPL-3.0-or-later license.
• This ensures that improvements remain open-source and benefit the community.

📖 Full license text: GNU AGPL v3.0