
Journal of Integrative Agriculture, Год журнала: 2024, Номер unknown
Опубликована: Дек. 1, 2024
Язык: Английский
Journal of Integrative Agriculture, Год журнала: 2024, Номер unknown
Опубликована: Дек. 1, 2024
Язык: Английский
Animals, Год журнала: 2024, Номер 14(8), С. 1226 - 1226
Опубликована: Апрель 19, 2024
In response to the high breakage rate of pigeon eggs and significant labor costs associated with egg-producing farming, this study proposes an improved YOLOv8-PG (real versus fake egg detection) model based on YOLOv8n. Specifically, Bottleneck in C2f module YOLOv8n backbone network neck are replaced Fasternet-EMA Block Fasternet Block, respectively. The is designed PConv (Partial Convolution) reduce parameter count computational load efficiently. Furthermore, incorporation EMA (Efficient Multi-scale Attention) mechanism helps mitigate interference from complex environments pigeon-egg feature-extraction capabilities. Additionally, Dysample, ultra-lightweight effective upsampler, introduced into further enhance performance lower overhead. Finally, EXPMA (exponential moving average) concept employed optimize SlideLoss propose EMASlideLoss classification loss function, addressing issue imbalanced data samples enhancing model's robustness. experimental results showed that F1-score, mAP50-95, mAP75 increased by 0.76%, 1.56%, 4.45%, respectively, compared baseline model. Moreover, reduced 24.69% 22.89%, Compared detection models such as Faster R-CNN, YOLOv5s, YOLOv7, YOLOv8s, exhibits superior performance. reduction contributes lowering deployment facilitates its implementation mobile robotic platforms.
Язык: Английский
Процитировано
10Plants, Год журнала: 2025, Номер 14(3), С. 365 - 365
Опубликована: Янв. 25, 2025
Efficient and accurate apple detection is crucial for the operation of apple-picking robots. To improve accuracy speed, we propose a lightweight apple-detection model based on YOLOv8n framework. The proposed introduces novel Self-Calibrated Coordinate (SCC) attention module, which enhances feature extraction, especially partially occluded apples, by effectively capturing spatial channel information. Additionally, replace C2f module within neck with Partial Convolution Module improved Reparameterization (PCMR), accelerates detection, reduces redundant computations, minimizes both parameter count memory access during inference. further optimize model, fuse multi-scale features from second third pyramid levels backbone architecture, achieving design suitable real-time detection. address missed detections misclassifications, Polynomial Loss (PolyLoss) integrated, enhancing class discrimination different subcategories. Compared to original YOLOv8n, increases mAP 2.90% 88.90% improves speed 220 FPS, 30.55% faster. it 89.36% FLOPs 2.47%. Experimental results demonstrate that outperforms mainstream object-detection algorithms, including Faster R-CNN, RetinaNet, SSD, RT-DETR-R18, RT-DETR-R34, YOLOv5n, YOLOv6-N, YOLOv7-tiny, YOLOv9-T YOLOv11n, in speed. Notably, has been used develop an Android application deployed iQOO Neo6 SE smartphone, 40 FPS 26.93% improvement over corresponding deployment enabling This study provides valuable reference designing efficient models resource-constrained
Язык: Английский
Процитировано
1Smart Agricultural Technology, Год журнала: 2025, Номер unknown, С. 100829 - 100829
Опубликована: Фев. 1, 2025
Язык: Английский
Процитировано
1Research Square (Research Square), Год журнала: 2024, Номер unknown
Опубликована: Авг. 8, 2024
Язык: Английский
Процитировано
3Опубликована: Янв. 1, 2025
Язык: Английский
Процитировано
0Agronomy, Год журнала: 2025, Номер 15(3), С. 582 - 582
Опубликована: Фев. 26, 2025
The conventional hand-picking of tea buds is inefficient and leads to inconsistent quality. Innovations in bud identification automated grading are essential for enhancing industry competitiveness. Key breakthroughs include detection accuracy lightweight model deployment. Traditional image recognition struggles with variable weather conditions, while high-precision models often too bulky mobile applications. This study proposed a YOLOV5 model, which was tested on three types across different scenarios. It incorporated convolutional network compact feature extraction layer, significantly reduced parameter computation. achieved 92.43% precision 87.25% mean average (mAP), weighing only 4.98 MB improving by 6.73% 2.11% reducing parameters 2 141.02 compared YOLOV5n6 YOLOV5l6. Unlike networks that detected single or dual grades, this offered refined advantages both size, making it suitable embedded devices limited resources. Thus, the YOLOV5n6_MobileNetV3 enhanced supported intelligent harvesting research technology.
Язык: Английский
Процитировано
0Measurement, Год журнала: 2025, Номер unknown, С. 117321 - 117321
Опубликована: Март 1, 2025
Язык: Английский
Процитировано
0Smart Agricultural Technology, Год журнала: 2025, Номер unknown, С. 100963 - 100963
Опубликована: Апрель 1, 2025
Язык: Английский
Процитировано
0Measurement, Год журнала: 2024, Номер 238, С. 115388 - 115388
Опубликована: Июль 24, 2024
Язык: Английский
Процитировано
2Plant Phenomics, Год журнала: 2024, Номер 6
Опубликована: Янв. 1, 2024
Our research focuses on winter jujube trees and is conducted in a greenhouse environment structured orchard to effectively control various growth conditions. The development of robotic system for harvesting crucial achieving mechanized harvesting. Harvesting jujubes efficiently requires accurate detection location. To address this issue, we proposed localization method based the MobileVit-Large selective kernel-GSConv-YOLO (MLG-YOLO) model. First, dataset constructed comprise scenarios lighting conditions leaf obstructions train Subsequently, MLG-YOLO model YOLOv8n proposed, with improvements including incorporation MobileViT reconstruct backbone keep more lightweight. neck enhanced LSKblock capture broader contextual information, lightweight convolutional technology GSConv introduced further improve accuracy. Finally, 3-dimensional combining RGB-D cameras proposed. Through ablation studies, comparative experiments, error tests, full-scale tree tests laboratory environments environments, effectiveness detecting locating confirmed. With MLG-YOLO, mAP increases by 3.50%, while number parameters reduced 61.03% comparison baseline Compared mainstream object models, excels both accuracy size, 92.70%, precision 86.80%, recall 84.50%, size only 2.52 MB. average environmental testing reached 100%, 92.82%. absolute positioning errors
Язык: Английский
Процитировано
2