Comparing YOLOv8 and Mask R-CNN for instance segmentation in complex orchard environments

Artificial Intelligence in Agriculture, Journal Year: 2024, Volume and Issue: 13, P. 84 - 99

Published: July 17, 2024

Instance segmentation, an important image processing operation for automation in agriculture, is used to precisely delineate individual objects of interest within images, which provides foundational information various automated or robotic tasks such as selective harvesting and precision pruning. This study compares the one-stage YOLOv8 two-stage Mask R-CNN machine learning models instance segmentation under varying orchard conditions across two datasets. Dataset 1, collected dormant season, includes images apple trees, were train multi-object delineating tree branches trunks. 2, early growing canopies with green foliage immature (green) apples (also called fruitlet), single-object only apples. The results showed that performed better than R-CNN, achieving good near-perfect recall both datasets at a confidence threshold 0.5. Specifically, achieved 0.90 0.95 all classes. In comparison, demonstrated 0.81 same dataset. With 0.93 0.97. this single-class scenario, 0.85 0.88. Additionally, inference times 10.9 ms multi-class (Dataset 1) 7.8 2), compared 15.6 12.8 by R-CNN's, respectively. These findings show YOLOv8's superior accuracy efficiency applications models, specifically Mask-R-CNN, suggests its suitability developing smart operations, particularly when real-time are necessary cases fruit thinning.

Language: Английский

---

A Seamless Deep Learning Approach for Apple Detection, Depth Estimation, and Tracking Using YOLO Models Enhanced by Multi-Head Attention Mechanism

Computers, Journal Year: 2024, Volume and Issue: 13(3), P. 83 - 83

Published: March 21, 2024

Considering precision agriculture, recent technological developments have sparked the emergence of several new tools that can help to automate agricultural process. For instance, accurately detecting and counting apples in orchards is essential for maximizing harvests ensuring effective resource management. However, there are intrinsic difficulties with traditional techniques identifying orchards. To identify, recognize, detect apples, apple target detection algorithms, such as YOLOv7, shown a great deal reflection accuracy. But occlusions, electrical wiring, branches, overlapping pose severe issues precisely apples. Thus, overcome these recognize find depth from drone-based videos complicated backdrops, our proposed model combines multi-head attention system YOLOv7 object identification framework. Furthermore, we provide ByteTrack method real time, which guarantees monitoring verify efficacy suggested model, thorough comparison assessment performed current techniques. The outcomes adequately proved effectiveness strategy, continuously surpassed competing methods achieve exceptional accuracies 0.92, 0.96, 0.95 respect precision, recall, F1 score, low MAPE 0.027, respectively.

Language: Английский

---

Real-time lettuce-weed localization and weed severity classification based on lightweight YOLO convolutional neural networks for intelligent intra-row weed control

Computers and Electronics in Agriculture, Journal Year: 2024, Volume and Issue: 226, P. 109404 - 109404

Published: Sept. 4, 2024

Language: Английский

---

MAE-YOLOv8-based small object detection of green crisp plum in real complex orchard environments

Computers and Electronics in Agriculture, Journal Year: 2024, Volume and Issue: 226, P. 109458 - 109458

Published: Sept. 20, 2024

Language: Английский

---

Drone-based apple detection: Finding the depth of apples using YOLOv7 architecture with multi-head attention mechanism

Smart Agricultural Technology, Journal Year: 2023, Volume and Issue: 5, P. 100311 - 100311

Published: Aug. 24, 2023

The agriculture drones are flocking and hovering over the crop fields to collect data or perform tasks related management. rise of artificial intelligence machine learning algorithms paved path innovative approaches in agriculture. Apple detection on farms has been a distinguished area expertise. apple target recognition like YOLOv7 gained lot reflection accuracy identify, recognize detect apples. However, significant problems with accurate time apples include occlusions, wiring, branching, overlapping. So, overcome this problem, deep approach model is projected rectify margin error drone-based inference live field. Along apples, depth from drone offers valued for optimizing harvesting, assessing yield, discovering diseases, handling orchards, evolving agricultural research industry. A specific multi-head attention mechanism applied capture spatial channel-wise dependencies concurrently. It can help complex interactions between regions features, improving accuracy. designed backgrounds better. identifies minimal objects enhances quality features achieve bounding boxes, which maximizes Incorporating function evaluate loss further increases model's According comparative study, proposed using modified Yolov7 architecture attains good 0.91, 0.96, 0.92 concerning precision, recall, F1-score, respectively.

Language: Английский

---

Optimizing the YOLOv7-Tiny Model with Multiple Strategies for Citrus Fruit Yield Estimation in Complex Scenarios

Agriculture, Journal Year: 2024, Volume and Issue: 14(2), P. 303 - 303

Published: Feb. 13, 2024

The accurate identification of citrus fruits is important for fruit yield estimation in complex orchards. In this study, the YOLOv7-tiny-BVP network constructed based on YOLOv7-tiny network, with as research object. This introduces a BiFormer bilevel routing attention mechanism, which replaces regular convolution GSConv, adds VoVGSCSP module to neck and simplified efficient layer aggregation (ELAN) partial (PConv) backbone network. improved model significantly reduces number parameters inference time, while maintaining network’s high recognition rate fruits. results showed that accuracy modified was 97.9% test dataset. Compared YOLOv7-tiny, size were reduced by 38.47% 4.6 MB, respectively. Moreover, accuracy, frames per second (FPS), F1 score 0.9, 2.02, 1%, proposed paper has an even after are 38.47%, only 7.7 provides new idea development lightweight target detection model.

Language: Английский

---

Precision agriculture with YOLO-Leaf: advanced methods for detecting apple leaf diseases

Frontiers in Plant Science, Journal Year: 2024, Volume and Issue: 15

Published: Oct. 15, 2024

The detection of apple leaf diseases plays a crucial role in ensuring crop health and yield. However, due to variations lighting shadow, as well the complex relationships between perceptual fields target scales, current methods face significant challenges. To address these issues, we propose new model called YOLO-Leaf. Specifically, YOLO-Leaf utilizes Dynamic Snake Convolution (DSConv) for robust feature extraction, employs BiFormer enhance attention mechanism, introduces IF-CIoU improve bounding box regression increased accuracy generalization ability. Experimental results on FGVC7 FGVC8 datasets show that significantly outperforms existing models terms accuracy, achieving mAP50 scores 93.88% 95.69%, respectively. This advancement not only validates effectiveness our approach but also highlights its practical application potential agricultural disease detection.

Language: Английский

---

An efficient method for detecting Camellia oleifera fruit under complex orchard environment

Scientia Horticulturae, Journal Year: 2024, Volume and Issue: 330, P. 113091 - 113091

Published: March 13, 2024

Language: Английский

---

MSOAR-YOLOv10: Multi-Scale Occluded Apple Detection for Enhanced Harvest Robotics

Horticulturae, Journal Year: 2024, Volume and Issue: 10(12), P. 1246 - 1246

Published: Nov. 25, 2024

The accuracy of apple fruit recognition in orchard environments is significantly affected by factors such as occlusion and lighting variations, leading to issues missed false detections. To address these challenges, particularly related occluded apples, this study proposes an improved apple-detection model, MSOAR-YOLOv10, based on YOLOv10. Firstly, a multi-scale feature fusion network enhanced adding 160 × scale layer the backbone network, which increases model’s sensitivity small local features, for fruits. Secondly, Squeeze-and-Excitation (SE) attention mechanism integrated into C2fCIB convolution module improve network’s focus regions interest input images. Additionally, Diverse Branch Block (DBB) introduced enhance performance convolutional neural network. Furthermore, Normalized Wasserstein Distance (NWD) loss function proposed effectively reduce detections densely packed overlapping targets. Experimental results orchards indicate that YOLOv10 model achieves precision, recall, mean average precision rates 89.3%, 89.8%, 92.8%, respectively, representing 3.1%, 2.2%, 3.0% compared original model. These validate enhances complex environments, improving operational harvesting robots real-world conditions.

Language: Английский

---

Optimizing precision agriculture: A real-time detection approach for grape vineyard unhealthy leaves using deep learning improved YOLOv7 with feature extraction capabilities

Computers and Electronics in Agriculture, Journal Year: 2025, Volume and Issue: 231, P. 109969 - 109969

Published: Jan. 30, 2025

Language: Английский

---