AMIGA_fruit_counting

Fruit detection and tracking for yield estimation using AMIGA robot.

Introduction

This project is a python implementation for fruit detection and tracking using YOLOv5 and ByteTrack. The project was build to be tested jointly with AMIGA robot for in-field fruit conting purposes. It was tested in the Farm@thon organized by Lleida Drone, , where it was recognized with the first prize. The code was prepared to be used with a ZED camera, but it could be easily adapted for other cameras such as Oak-D.

Preparation

First of all, create a new project folder and clone the code inside:

mkdir new_project
cd new_project
git clone https://github.com/GRAP-UdL-AT/AMIGA_fruit_counting.git

Then, install the project requirements:

cd AMIGA_fruit_counting
pip install -r requirements.txt

Clone the YOLOv5 repository and install it following the instructions at YOLOv5 repository:

cd yolov5
git clone https://github.com/ultralytics/yolov5.git

Data Preparation

Make a new folder were yolo weights will be saved:

cd ..
cd ..
mkdir yolo_weights

Then, train a YOLO model or download the following pretrained weights.

Launch the code

• Execute the file /new_project/AMIGA_fruit_counting/zed_fruit_counting.py

Authorship

This project is contributed by Francesc Net Barnes, Marc Felip Pomés and Jordi Gené-Mola from GRAP-UdL-AT.

Please contact authors to report bugs @ jordi.genemola@udl.cat

Acknowledgements

This work was partly funded by the Spanish Ministry of Science, Innovation and Universities (grant RTI2018-094222-B-I00[PAgFRUIT project] by MCIN/AEI/10.13039/501100011033 and by “ERDF, a way of making Europe”, by the European Union).