🌴 PalmsCNN: Palm Tree Detection RPAs (QGIS Plugin)

PalmsCNN is a QGIS plugin that applies Deep Learning models to automatically detect three ecologically and economically important Amazonian palm species:
aguaje/buriti (Mauritia flexuosa), huasai/acai (Euterpe precatoria), and ungurahui/patawa (Oenocarpus bataua)

More information about the models can be found in Tagle et al., 2025, Nature Communications.

This plugin was developed within the framework of the projects “Supervisiones Optimizadas” and “New approaches to understand the state of biodiversity and contribute to social well-being: studying the distribution and degradation of Mauritia flexuosa in the Amazon”, through the collaboration of OSINFOR, IIAP, SERNANP, the University of Leeds, the University of Brescia, and Wageningen University. And with financial support from Newton Fund, FONDECYT, WWF, GIZ, and USAID.

---

🧠 Overview

PalmsCNN uses Convolutional Neural Networks (CNNs) exported to the ONNX format to recognize individual palm crowns in RGB orthomosaic images captured by drones or high-resolution satellites.
The plugin enables automated, reproducible, and cost-efficient mapping of Amazonian palm ecosystems.

---

🌴 Target Palm Species

Scientific name	Common name (Spanish / Portuguese)
Mauritia flexuosa	aguaje/ buriti
Euterpe precatoria	huasai/ acai
Oenocarpus bataua	ungurahui/ patawa

---

🏗️ Plugin Architecture

---

⚙️ Key Features

• Automatic palm detection from RPAs RGB imagery (no multispectral data required, nor canopy heights, only simple RGB images).
• CNN models in ONNX format.
• Full integration with the QGIS Processing Toolbox.
• Georeferenced output layers (vector or raster).
• Cross-platform support (Windows, Linux, macOS).
• Automatic setup of a Python virtual environment (venv) for dependencies.

---

🧩 Inputs and Outputs

Inputs

• Input Raster
  Select an RGB orthomosaic image in .tif format.
  This georeferenced image serves as the main input for palm detection and classification.

• Output Folder and Filename
  Specify the folder path and name for the output georeferenced classified raster.
  This folder will also serve as the working directory for all generated outputs.

Outputs

• Output Raster
  A georeferenced classified image showing the detected palm crowns labeled by species.

• Output Vector (Shapefile)
  A vector layer (.shp) containing polygons for each detected palm crown.

• Centroid Layer
  A point vector layer showing the centroid (center coordinates) of each detected palm.

• Attributes Table (.csv)
  A table containing detailed information for each detected palm:

  • id → Unique palm identifier
  • class_species → Predicted species (Mauritia flexuosa, Euterpe precatoria, Oenocarpus bataua)
  • area_m2 → Area of the palm crown (in square meters)
  • utm_x, utm_y → UTM coordinates of the palm centroid

• Summary Report (.csv)
  Summary statistics including:

  • Number of detected palms per species
  • Total area (m²) occupied by species
  • Overall total number of detected palms

---

📦 Installation

🔹 From ZIP (for end users)

1. Download the latest release from Releases.
2. In QGIS, open:
   Plugins → Manage and Install Plugins → Install from ZIP.
3. Select the file:
   deteccion_de_palmeras-<version>.zip
4. Click Install Plugin.

🔹 From source (for developers)

git clone https://github.com/iiap-gob-pe/PalmsCNN-plugin-QGIS.git
d PalmsCNN-plugin-QGIS/deteccion_de_palmeras/help
make package        # or make.bat package on Windows

---

✅ User Manual

Please download the user manual by clicking the provided link.

📊 Sample Data

The test data for testing can be downloaded by clicking the provided link.

🌍 Credits

Co-developed by IIAP, OSINFOR, SERNANP, University of Brescia, Wageningen University and University of Leeds within the framework of the projects “Supervisiones Optimizadas” and “New approaches to understand the state of biodiversity and contribute to social well-being.”

Funding provided by Newton Fund, Embajada Britanica Lima, FONDECYT PERU, WWF - Russel E. Train Education for Nature Programme (EFN), GIZ, and USAID.

Maintained by the Instituto de Investigaciones de la Amazonía Peruana (IIAP) Laboratorio de Inteligencia Artificial - Programa BOSQUES Iquitos, Peru

---

🧾 License

This project is distributed under the MIT License.

© 2025 Instituto de Investigaciones de la Amazonía Peruana (IIAP). Free for scientific, educational, and conservation use.

---

🔗 References

• Palacios, S. (2020). Aguaje QGIS plugin: Tool for detecting Mauritia flexuosa (Aguaje) palms in raster aerial images (Master’s thesis). University of Brescia, Italy.
• Tagle Casapia, X., Cardenas-Vigo, R., Marcos, D. et al. (2025) Effective integration of drone technology for mapping and managing palm species in the Peruvian Amazon. Nature Communications. https://doi.org/10.1038/s41467-025-58358-5
• QGIS Documentation — https://docs.qgis.org
• ONNX Runtime — https://onnxruntime.ai
• IIAP — https://www.iiap.gob.pe

---

📚 How to Cite

If you use the QGIS plugin Palms Detection RPAs, please cite:

Palacios, S., Tagle, X, Falen, L., Di Liberto, S., Minhuey. A., Torres, S., Baker, T., Fernandez, E., Allcahuaman, E., Campos, L., Adami, N., Signoroni, A. Cárdenas, R. (in prep). Stakeholder driven Development of a Deep Learning-Based QGIS Plugin for Identifying Palm Trees in Tropical Forests Available at: https://github.com/iiap-gob-pe/PalmsCNN-plugin-QGIS Contact: rcardenasv@iiap.gob.pe

If you use the model PalmsCNN, please cite:

Tagle Casapia, X., Cardenas-Vigo, R., Marcos, D. et al. (2025) Effective integration of drone technology for mapping and managing palm species in the Peruvian Amazon. Nature Communications. https://doi.org/10.1038/s41467-025-58358-5