Abstract
Accurate quantification of the citrus dropped number plays a vital role in evaluating the disaster resistance capabilities of citrus varieties and selecting superior cultivars. However, research in this critical area remains notably insufficient. To bridge this gap, we conducted in-depth experiments using a custom dataset of 1200 citrus images and proposed a lightweight YOLO-MECD model that is built upon the YOLOv11s architecture. Firstly, the EMA attention mechanism was introduced as a replacement for the traditional C2PSA attention mechanism. This modification not only enhances feature extraction capabilities and detection accuracy for citrus fruits but also achieves a significant reduction in model parameters. Secondly, we implemented a CSPPC module based on partial convolution to replace the original C3K2 module, effectively reducing both parameter count and computational complexity while maintaining mAP values. At last, the MPDIoU loss function was employed, resulting in improved bounding box detection accuracy and accelerated model convergence. Notably, our research reveals that reducing convolution operations in the backbone architecture substantially enhances small object detection capabilities and significantly decreases model parameters, proving more effective than the addition of small object detection heads. The experimental results and comparative analysis with similar network models indicate that the YOLO-MECD model has achieved significant improvements in both detection performance and computational efficiency. This model demonstrates excellent comprehensive performance in citrus object detection tasks, with a precision (P) of 84.4%, a recall rate (R) of 73.3%, and an elevated mean average precision (mAP) of 81.6%. Compared to the baseline, YOLO-MECD has improved by 0.2, 4.1, and 3.9 percentage points in detection precision, recall rate, and mAP value, respectively. Furthermore, the number of model parameters has been substantially reduced from 9,413,574 in YOLOv11s to 2,297,334 (a decrease of 75.6%), and the model size has been compressed from 18.2 MB to 4.66 MB (a reduction of 74.4%). Moreover, YOLO-MECD also demonstrates superior performance against contemporary models, with mAP improvements of 3.8%, 3.2%, and 5.5% compared to YOLOv8s, YOLOv9s, and YOLOv10s, respectively. The model’s versatility is evidenced by its excellent detection performance across various citrus fruits, including pomelos and kumquats. These achievements establish YOLO-MECD as a robust technical foundation for advancing citrus fruit detection systems and the development of smart orchards.
Abstract from the paper’s OpenAlex record; open copy licensed cc-by. Read at the source ↗
The text the model read, with the source record’s inline markup (italics, sub- and superscripts) rendered and its character entities decoded; quote offsets refer to the raw text.
Claims 16
Claim · model-written summary
Replacing C2PSA with EMA improved citrus detection accuracy and reduced model parameters.
CitrusNot reportedComparatorValue
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e0“This modification not only enhances feature extraction capabilities and detection accuracy for citrus fruits but also achieves a significant reduction in model parameters”
Claim · model-written summary
CSPPC reduced parameter count and computational complexity while maintaining mAP.
Not reportedTaxonComparatorValue
Verbatim
e0“effectively reducing both parameter count and computational complexity while maintaining mAP values”
Claim · model-written summary
MPDIoU improved bounding-box accuracy and convergence.
Not reportedTaxonComparatorValue
Verbatim
e0“resulting in improved bounding box detection accuracy and accelerated model convergence”
Claim · model-written summary
Reducing backbone convolutions improved small-object detection and decreased parameters.
Not reportedTaxonComparatorValue
Verbatim
e0“reducing convolution operations in the backbone architecture substantially enhances small object detection capabilities and significantly decreases model parameters”
Claim · model-written summary
YOLO-MECD precision was 84.4%, 0.2 points above baseline.
Not reportedTaxonDirection
Verbatim
e0“with a precision (P) of 84.4%, a recall rate (R) of 73.3%, and an elevated mean average precision (mAP) of 81.6%”
Claim · model-written summary
YOLO-MECD recall was 73.3%.
Not reportedTaxonDirection
Verbatim
e0“a recall rate (R) of 73.3%”
Claim · model-written summary
YOLO-MECD mAP was 81.6%.
Not reportedTaxon
Verbatim
e0“elevated mean average precision (mAP) of 81.6%”
Claim · model-written summary
Against baseline, precision improved 0.2 percentage points.
Not reportedTaxon
Verbatim
e0“Compared to the baseline, YOLO-MECD has improved by 0.2, 4.1, and 3.9 percentage points in detection precision, recall rate, and mAP value, respectively”
Claim · model-written summary
Against baseline, recall improved 4.1 percentage points.
Not reportedTaxon
Verbatim
e0“Compared to the baseline, YOLO-MECD has improved by 0.2, 4.1, and 3.9 percentage points in detection precision, recall rate, and mAP value, respectively”
Claim · model-written summary
Against baseline, mAP improved 3.9 percentage points.
Not reportedTaxon
Verbatim
e0“Compared to the baseline, YOLO-MECD has improved by 0.2, 4.1, and 3.9 percentage points in detection precision, recall rate, and mAP value, respectively”
Claim · model-written summary
Model parameters fell from 9,413,574 to 2,297,334, a 75.6% decrease.
- Outcome
- number of model parameterse0
- Direction
- reducede0
- Value
- 2,297,334e0
- Qualifier
- a decrease of 75.6%e0
- Comparator
- YOLOv11se0
Not reportedTaxon
Verbatim
e0“number of model parameters has been substantially reduced from 9,413,574 in YOLOv11s to 2,297,334 (a decrease of 75.6%)”
Claim · model-written summary
Model size fell from 18.2 MB to 4.66 MB, a 74.4% reduction.
Not reportedTaxonComparator
Verbatim
e0“model size has been compressed from 18.2 MB to 4.66 MB (a reduction of 74.4%)”
Claim · model-written summary
mAP exceeded YOLOv8s by 3.8%.
Not reportedTaxon
Verbatim
e0“mAP improvements of 3.8%, 3.2%, and 5.5% compared to YOLOv8s, YOLOv9s, and YOLOv10s, respectively”
Claim · model-written summary
mAP exceeded YOLOv9s by 3.2%.
Not reportedTaxon
Verbatim
e0“mAP improvements of 3.8%, 3.2%, and 5.5% compared to YOLOv8s, YOLOv9s, and YOLOv10s, respectively”
Claim · model-written summary
mAP exceeded YOLOv10s by 5.5%.
Not reportedTaxon
Verbatim
e0“mAP improvements of 3.8%, 3.2%, and 5.5% compared to YOLOv8s, YOLOv9s, and YOLOv10s, respectively”
Claim · model-written summary
The model detected pomelos and kumquats well.
- Outcome
- detection performancee0
Not reportedComparatorDirectionValue
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e0“excellent detection performance across various citrus fruits, including pomelos and kumquats”