An Open-Source Pipeline for UAV (Unmanned Aerial Vehicle)-Based Object Detection and Data Extraction
- Worasit Sangjan
- Piyush Pandey
- Norman B. Best
- Jacob D. Washburn
USDA-ARS, Plant Genetics Research Unit, Columbia, MO, United States
MatchPlant is an open-source pipeline designed for the automated detection of individual objects using UAV-derived imagery. The pipeline combines interactive tools for preparing UAV imagery with automated deep-learning methods and data extraction capabilities. For the case study, it is used for individual maize detection. By leveraging the Faster R-CNN object detection model trained on high-resolution, undistorted UAV images. MatchPlant also provides utilization modules to incorporate the detected bounding boxes and extract digital plant traits from an orthomosaic.
- Modular Design: Flexible framework adaptable to various agricultural applications
- Automated Object Detection: Uses Faster R-CNN for reliable individual plant detection
- High-Precision Processing: Trains on high-resolution, undistorted UAV images to avoid orthomosaic artifacts
- Geospatial Transformation: Accurately projects detected plant coordinates onto orthomosaic
- User-Friendly Tools: User-friendly GUI tools for data preparation and manual annotation
- Data Extraction & Analysis: Enables spatial analysis and phenotypic trait extraction
This repository is organized into the following modules, each containing specific components of the pipeline:
-
Data Preprocessing: Prepare and optimize UAV imagery
- GPS data embedding tool
- GCP file creation tool
- Orthomosaic generation using OpenDroneMap
-
Data Preparation: Label and organize training, validation, and testing data
- Optimal UAV image dataset creation tool
- Interactive labeling tool
- Data tiling and splitting tool
-
Model Development: Train and test detection models, including the transfer learning model
- Faster R-CNN training and validation
- Model testing
- Transfer learning utility
-
Utilization: Project location and create a layer to extract object features
- Detection projection tool (Module 8 is under improvement!)
- Shapefile generation tool
Each module contains its detailed README with specific installation instructions and usage guidelines.
Figure 1: Diagram of the MathPlant modular open-source pipeline
1-GPS: Global Position System, 2-UAV: Unmanned Aerial Vehicle, 3-GCP: Ground Control Point, 4-GUI: Graphical User Interface, 5-COCO: Common Objects in Context, 6-YOLO: You Only Look Once, 7-Faster R-CNN: Faster Region-based Convolutional Neural Network
- Python 3.9+
- OpenDroneMap (ODM)
- Additional requirements listed in module-specific documentation
To begin using MatchPlant:
- Choose the appropriate module for your task
- Follow the module-specific installation instructions
- Refer to the module README for detailed usage guidelines
To use the MatchPlant pipeline with our prepared dataset, download from Zenodo:
Zenodo Repository: https://zenodo.org/records/14856123
Dataset Contents:
- UAV images: Undistorted images created by OpenDroneMap software from high-resolution RGB images collected during the 2021 growing season
- Annotation file: COCO format bounding box annotations (.json files) for individual maize detection
- Pre-trained model: Faster R-CNN model trained on the UAV images for individual maize detection (use in 6-2_obj_det_trans_learner)
Download Options:
- Complete Dataset Download
# Using wget wget https://zenodo.org/records/14856123/files/UAV%20images.zip wget https://zenodo.org/records/14856123/files/Annotation%20file.zip wget https://zenodo.org/records/14856123/files/Pre-trained%20model_Faster%20R-CNN.pt # Using curl curl -O https://zenodo.org/records/14856123/files/UAV%20images.zip curl -O https://zenodo.org/records/14856123/files/Annotation%20file.zip curl -O https://zenodo.org/records/14856123/files/Pre-trained%20model_Faster%20R-CNN.pt
- Individual File Downloads
- UAV images (ZIP): Download UAV images.zip
- Annotation file (ZIP): Download Annotation file.zip
- Pre-trained model: Download Pre-trained model_Faster R-CNN.pt
Getting Started with the Dataset:
After downloading the dataset (UAV images and Annotation file), please start with the module 5_img_splitter to use our pipeline.
If you use MatchPlant in your research, please cite:
Sangjan, W., Pandey, P., Best, N. B., & Washburn, J. D. (2025). MatchPlant: An Open-Source Pipeline for UAV-Based Single-Plant Detection and Data Extraction. arXiv preprint arXiv:2506.12295. https://doi.org/10.48550/arXiv.2506.12295
For the dataset, please cite:
Sangjan, W., Pandey, P., Best, N. B., & Washburn, J. D. (2025). MatchPlant: An open-source pipeline for UAV-based single-plant detection and data extraction [Dataset]. Zenodo. https://doi.org/10.5281/zenodo.14856123
For questions and collaboration opportunities, please contact:
Jacob D. Washburn; Email: [email protected]
This research was supported in part by an appointment to the Agricultural Research Service (ARS) Research Participation Program administered by the Oak Ridge Institute for Science and Education (ORISE) through an interagency agreement between the U.S. Department of Energy (DOE) and the U.S. Department of Agriculture (USDA). ORISE is managed by ORAU under DOE contract number DE-SC0014664. This research used resources provided by the SCINet project and/or the AI Center of Excellence of the USDA Agricultural Research Service, ARS project numbers 0201-88888-003-000D and 0201-88888-002-000D. Funding was also provided by the United States Department of Agriculture, Agricultural Research Service, SCINet Postdoctoral Fellows Program. All opinions expressed in this publication are the author’s and do not necessarily reflect the policies and views of USDA, DOE, or ORAU/ORISE.
This project is licensed under the MIT License. For details, see the LICENSE file.