Optimizing the loss function for bounding box regression through scale smoothing
Deep learning technology is widely used in target detection in machine vision. However, existing regression loss functions used for training networks suffer from slow convergence and imprecise localization, hindering the realization of fast and accurate visual detection. To address this, the study p...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2024-11-01
|
| Series: | Ain Shams Engineering Journal |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2090447924004210 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1846164567708663808 |
|---|---|
| author | Ying-Jun Lei Bo-Yu Wang Yu-Tong Yang |
| author_facet | Ying-Jun Lei Bo-Yu Wang Yu-Tong Yang |
| author_sort | Ying-Jun Lei |
| collection | DOAJ |
| description | Deep learning technology is widely used in target detection in machine vision. However, existing regression loss functions used for training networks suffer from slow convergence and imprecise localization, hindering the realization of fast and accurate visual detection. To address this, the study proposes the smoothing adaptive intersection over union loss (SAIoU Loss), which adapts bounding box regression through scale smoothing. By analyzing the bounding box regression process, SAIoU Loss incorporates a center-of-mass distance penalty term to enhance prediction speed during box distance regression in the pre-training phase. Additionally, it integrates a corner point distance penalty term with adaptive weights to refine the similarity of predicted box shapes throughout regression. The experimental results demonstrate that SAIoU Loss achieves a 39.6 mAP in target detection model training on PASCAL VOC, marking a 3.39% improvement. It also records the highest result of 26.7% in medium-sized target detection, which represents a 9.43% improvement over IoU. In the VisDrone 2019 dataset, SAIoU Loss reaches a detection accuracy of 14.8 mAP, improving by 1.3 mAP compared to the Baseline. The SAIoU loss proposed in this study realizes efficient and highly accurate target detection. |
| format | Article |
| id | doaj-art-ae1b991d1a804d5bbf3243f9452cc6cd |
| institution | Kabale University |
| issn | 2090-4479 |
| language | English |
| publishDate | 2024-11-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Ain Shams Engineering Journal |
| spelling | doaj-art-ae1b991d1a804d5bbf3243f9452cc6cd2024-11-18T04:32:59ZengElsevierAin Shams Engineering Journal2090-44792024-11-011511103046Optimizing the loss function for bounding box regression through scale smoothingYing-Jun Lei0Bo-Yu Wang1Yu-Tong Yang2School of Mechanical Engineering, Jiangsu University, Zhenjiang 212013, Jiangsu Province, ChinaSchool of Management, Jiangsu University, Zhenjiang 212013, Jiangsu Province, ChinaSchool of Mechanical Engineering, Jiangsu University, Zhenjiang 212013, Jiangsu Province, China; Corresponding author.Deep learning technology is widely used in target detection in machine vision. However, existing regression loss functions used for training networks suffer from slow convergence and imprecise localization, hindering the realization of fast and accurate visual detection. To address this, the study proposes the smoothing adaptive intersection over union loss (SAIoU Loss), which adapts bounding box regression through scale smoothing. By analyzing the bounding box regression process, SAIoU Loss incorporates a center-of-mass distance penalty term to enhance prediction speed during box distance regression in the pre-training phase. Additionally, it integrates a corner point distance penalty term with adaptive weights to refine the similarity of predicted box shapes throughout regression. The experimental results demonstrate that SAIoU Loss achieves a 39.6 mAP in target detection model training on PASCAL VOC, marking a 3.39% improvement. It also records the highest result of 26.7% in medium-sized target detection, which represents a 9.43% improvement over IoU. In the VisDrone 2019 dataset, SAIoU Loss reaches a detection accuracy of 14.8 mAP, improving by 1.3 mAP compared to the Baseline. The SAIoU loss proposed in this study realizes efficient and highly accurate target detection.http://www.sciencedirect.com/science/article/pii/S2090447924004210Object detectionLoss functionCorner point distanceCentre of mass distance |
| spellingShingle | Ying-Jun Lei Bo-Yu Wang Yu-Tong Yang Optimizing the loss function for bounding box regression through scale smoothing Ain Shams Engineering Journal Object detection Loss function Corner point distance Centre of mass distance |
| title | Optimizing the loss function for bounding box regression through scale smoothing |
| title_full | Optimizing the loss function for bounding box regression through scale smoothing |
| title_fullStr | Optimizing the loss function for bounding box regression through scale smoothing |
| title_full_unstemmed | Optimizing the loss function for bounding box regression through scale smoothing |
| title_short | Optimizing the loss function for bounding box regression through scale smoothing |
| title_sort | optimizing the loss function for bounding box regression through scale smoothing |
| topic | Object detection Loss function Corner point distance Centre of mass distance |
| url | http://www.sciencedirect.com/science/article/pii/S2090447924004210 |
| work_keys_str_mv | AT yingjunlei optimizingthelossfunctionforboundingboxregressionthroughscalesmoothing AT boyuwang optimizingthelossfunctionforboundingboxregressionthroughscalesmoothing AT yutongyang optimizingthelossfunctionforboundingboxregressionthroughscalesmoothing |