Particle swarm optimization with YOLOv8 for improved detection performance of tomato plants
Abstract Identification and precise classification of plants are crucial in improving plant quality and economic viability, particularly in an industrial setting. In a faster-growing world, there is a growing demand for fully automated tomato detection and grading systems. Within the past few years,...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
SpringerOpen
2025-06-01
|
| Series: | Journal of Big Data |
| Subjects: | |
| Online Access: | https://doi.org/10.1186/s40537-025-01206-6 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Identification and precise classification of plants are crucial in improving plant quality and economic viability, particularly in an industrial setting. In a faster-growing world, there is a growing demand for fully automated tomato detection and grading systems. Within the past few years, applying deep learning in detecting and classifying tomatoes into different classes has gained popularity. This study aims to build a new framework for the efficient automated harvesting of tomato plants based on deep learning. The new model integrates the capabilities of Particle Swarm Optimization (PSO) and the You Only Look Once version-8 (YOLOv8) architecture for better hyperparameter optimization and improved performance results. For the first time, it classifies and detects ripe, semi-ripe, and unripe tomatoes, in addition to diseased and rotten tomatoes. To validate the efficacy of the proposed YOLO-v8 network’s performance, three experiments were conducted employing a unique dataset, collected from different sources. Firstly, two experiments were conducted to formally confirm whether or not the utilized data augmentation technique significantly improved, one with data augmentation and another one with the end-to-end framework without data augmentation. Secondly, the proposed YOLOv8 was compared with other YOLO versions, e.g., YOLOv3, YOLOv5, S-YOLOv5, YOLOv7, and YOLOv8. Thirdly, the proposed framework was compared with many cutting-edge object detection architectures for tomato harvesting, e.g., Convolutional neural network (CNN), Mask R-CNN and color analysis, and other models based on handcrafted features. The enhancements made to the original YOLO-v8 network have attained promising results. The experiments on the collection of different datasets reveal that the proposed model performs with the highest precision, recall, F1-score, and mean average precision (mAP) of 0.89, 0.9, 0.89, and 0.89 (mAP@0.5:0.95), respectively, exceeding other models. This framework offers a viable and useful solution for class detection and location identification of tomato plants. |
|---|---|
| ISSN: | 2196-1115 |