BFE-Net: Bidirectional Multi-Scale Feature Enhancement for Small Object Detection
Small object detection becomes a challenging problem in computer vision due to low resolution and less feature information. Making full use of high-resolution features is an important factor in improving small object detection. In this paper, to improve the utilization of high-resolution features, t...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-04-01
|
| Series: | Applied Sciences |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2076-3417/12/7/3587 |
| _version_ | 1797440315081097216 |
|---|---|
| author | Qian Zhang Jie Ren Hong Liang Ying Yang Lu Chen |
| author_facet | Qian Zhang Jie Ren Hong Liang Ying Yang Lu Chen |
| author_sort | Qian Zhang |
| collection | DOAJ |
| description | Small object detection becomes a challenging problem in computer vision due to low resolution and less feature information. Making full use of high-resolution features is an important factor in improving small object detection. In this paper, to improve the utilization of high-resolution features, this work proposes the Bidirectional Multi-scale Feature Enhancement Network (BFE-Net) based on RetinaNet. First, this work introduces a bidirectional feature pyramid structure to shorten the propagation path of high-resolution features. Then, this work utilizes residually connected dilated convolutional blocks to fully extract high-resolution features of low-feature layers. Finally, this work supplements the high-resolution features lost in the high-level feature propagation process by leveraging the high-level guided lower-level features. Experiments show that our proposed BFE-Net achieves stable performance gains in the object detection task. Specifically, the improved method improves RetinaNet from 34.4 AP to 36.3 AP on the challenging MS COCO dataset and especially achieves excellent results in small object detection with an improvement of 2.8%. |
| first_indexed | 2024-03-09T12:05:25Z |
| format | Article |
| id | doaj.art-252be8739075450bbea847ebe964f4e8 |
| institution | Directory Open Access Journal |
| issn | 2076-3417 |
| language | English |
| last_indexed | 2024-03-09T12:05:25Z |
| publishDate | 2022-04-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Applied Sciences |
| spelling | doaj.art-252be8739075450bbea847ebe964f4e82023-11-30T22:57:48ZengMDPI AGApplied Sciences2076-34172022-04-01127358710.3390/app12073587BFE-Net: Bidirectional Multi-Scale Feature Enhancement for Small Object DetectionQian Zhang0Jie Ren1Hong Liang2Ying Yang3Lu Chen4College of Computer Science and Technology, China University of Petroleum (East China), Qingdao 266580, ChinaCollege of Computer Science and Technology, China University of Petroleum (East China), Qingdao 266580, ChinaCollege of Computer Science and Technology, China University of Petroleum (East China), Qingdao 266580, ChinaCollege of Computer Science and Technology, China University of Petroleum (East China), Qingdao 266580, ChinaCollege of Computer Science and Technology, China University of Petroleum (East China), Qingdao 266580, ChinaSmall object detection becomes a challenging problem in computer vision due to low resolution and less feature information. Making full use of high-resolution features is an important factor in improving small object detection. In this paper, to improve the utilization of high-resolution features, this work proposes the Bidirectional Multi-scale Feature Enhancement Network (BFE-Net) based on RetinaNet. First, this work introduces a bidirectional feature pyramid structure to shorten the propagation path of high-resolution features. Then, this work utilizes residually connected dilated convolutional blocks to fully extract high-resolution features of low-feature layers. Finally, this work supplements the high-resolution features lost in the high-level feature propagation process by leveraging the high-level guided lower-level features. Experiments show that our proposed BFE-Net achieves stable performance gains in the object detection task. Specifically, the improved method improves RetinaNet from 34.4 AP to 36.3 AP on the challenging MS COCO dataset and especially achieves excellent results in small object detection with an improvement of 2.8%.https://www.mdpi.com/2076-3417/12/7/3587small object detectiondilated convolutionattention mechanismfeature pyramid |
| spellingShingle | Qian Zhang Jie Ren Hong Liang Ying Yang Lu Chen BFE-Net: Bidirectional Multi-Scale Feature Enhancement for Small Object Detection Applied Sciences small object detection dilated convolution attention mechanism feature pyramid |
| title | BFE-Net: Bidirectional Multi-Scale Feature Enhancement for Small Object Detection |
| title_full | BFE-Net: Bidirectional Multi-Scale Feature Enhancement for Small Object Detection |
| title_fullStr | BFE-Net: Bidirectional Multi-Scale Feature Enhancement for Small Object Detection |
| title_full_unstemmed | BFE-Net: Bidirectional Multi-Scale Feature Enhancement for Small Object Detection |
| title_short | BFE-Net: Bidirectional Multi-Scale Feature Enhancement for Small Object Detection |
| title_sort | bfe net bidirectional multi scale feature enhancement for small object detection |
| topic | small object detection dilated convolution attention mechanism feature pyramid |
| url | https://www.mdpi.com/2076-3417/12/7/3587 |
| work_keys_str_mv | AT qianzhang bfenetbidirectionalmultiscalefeatureenhancementforsmallobjectdetection AT jieren bfenetbidirectionalmultiscalefeatureenhancementforsmallobjectdetection AT hongliang bfenetbidirectionalmultiscalefeatureenhancementforsmallobjectdetection AT yingyang bfenetbidirectionalmultiscalefeatureenhancementforsmallobjectdetection AT luchen bfenetbidirectionalmultiscalefeatureenhancementforsmallobjectdetection |