SFGNet detecting objects via spatial fine-grained feature and enhanced RPN with spatial context
Object detection, which is one of the most fundamental visual recognition tasks, has been a hotspot in computer vision. CNN (Convolutional Neural Networks) have been widely employed for building detector. Due to the success of RPN (Region Proposal Network), the two-stage detectors get both classific...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Taylor & Francis Group
2022-12-01
|
| Series: | Systems Science & Control Engineering |
| Subjects: | |
| Online Access: | https://www.tandfonline.com/doi/10.1080/21642583.2022.2062479 |
| _version_ | 1818188438259630080 |
|---|---|
| author | Jun Hu Yongfeng Wang Shuai Cheng Jiaxin Liu Jiawen Kang Wenxing Yang |
| author_facet | Jun Hu Yongfeng Wang Shuai Cheng Jiaxin Liu Jiawen Kang Wenxing Yang |
| author_sort | Jun Hu |
| collection | DOAJ |
| description | Object detection, which is one of the most fundamental visual recognition tasks, has been a hotspot in computer vision. CNN (Convolutional Neural Networks) have been widely employed for building detector. Due to the success of RPN (Region Proposal Network), the two-stage detectors get both classification accuracy and precise regression bounding boxes. However, they still struggle in small-size object detection. In this paper, we present a deep network, namely Spatial Fine-Grained Network (SFGN). The SFGN that exploits Spatial Fine-Grained Features (SFGF) concatenates the higher resolution features, which is fine-grained with the low resolution features and high-level semantic by stacking spatial features for fine-grained features. An enhanced region proposal generator is proposed to get the objectless for small object to obtain a small set of proposal. The contextual information surrounding the region of interest is embedded using local spatial information for increasing the useful information and discriminating the background. For improving the detection performance, we use a simple yet surprisingly effective online hard example mining (OHEM) algorithm for training region proposal generator. It embeds an efficiently implemented soft non-maximum suppression (soft-NMS) for replacing with tradition NMS to obtain consistent improvements without increasing the computational complexity in inference. On PASCAL VOC 2007 and PASCAL VOC 2012 datasets, our SFGN improves baseline model from 81.2% mAP to 80.6% mAP. On MS COCO dataset, SFGN also performs better than baseline model. As intuition suggests, our detection results provide strong evidence that our SFGN improves detection accuracy, especially in small object test. |
| first_indexed | 2024-12-11T23:26:55Z |
| format | Article |
| id | doaj.art-6ffc9fc8834e43139739b57865b0487b |
| institution | Directory Open Access Journal |
| issn | 2164-2583 |
| language | English |
| last_indexed | 2024-12-11T23:26:55Z |
| publishDate | 2022-12-01 |
| publisher | Taylor & Francis Group |
| record_format | Article |
| series | Systems Science & Control Engineering |
| spelling | doaj.art-6ffc9fc8834e43139739b57865b0487b2022-12-22T00:46:09ZengTaylor & Francis GroupSystems Science & Control Engineering2164-25832022-12-0110138840610.1080/21642583.2022.2062479SFGNet detecting objects via spatial fine-grained feature and enhanced RPN with spatial contextJun Hu0Yongfeng Wang1Shuai Cheng2Jiaxin Liu3Jiawen Kang4Wenxing Yang5School of Computer Science and Engineering, Northeastern University, Liaoning, People's Republic of ChinaNeusoft Reachauto Corporation, Liaoning, People's Republic of ChinaNeusoft Reachauto Corporation, Liaoning, People's Republic of ChinaSchool of Computer Science and Engineering, Northeastern University, Liaoning, People's Republic of ChinaSchool of Computer Science and Engineering, Northeastern University, Liaoning, People's Republic of ChinaSchool of Computer Science and Engineering, Northeastern University, Liaoning, People's Republic of ChinaObject detection, which is one of the most fundamental visual recognition tasks, has been a hotspot in computer vision. CNN (Convolutional Neural Networks) have been widely employed for building detector. Due to the success of RPN (Region Proposal Network), the two-stage detectors get both classification accuracy and precise regression bounding boxes. However, they still struggle in small-size object detection. In this paper, we present a deep network, namely Spatial Fine-Grained Network (SFGN). The SFGN that exploits Spatial Fine-Grained Features (SFGF) concatenates the higher resolution features, which is fine-grained with the low resolution features and high-level semantic by stacking spatial features for fine-grained features. An enhanced region proposal generator is proposed to get the objectless for small object to obtain a small set of proposal. The contextual information surrounding the region of interest is embedded using local spatial information for increasing the useful information and discriminating the background. For improving the detection performance, we use a simple yet surprisingly effective online hard example mining (OHEM) algorithm for training region proposal generator. It embeds an efficiently implemented soft non-maximum suppression (soft-NMS) for replacing with tradition NMS to obtain consistent improvements without increasing the computational complexity in inference. On PASCAL VOC 2007 and PASCAL VOC 2012 datasets, our SFGN improves baseline model from 81.2% mAP to 80.6% mAP. On MS COCO dataset, SFGN also performs better than baseline model. As intuition suggests, our detection results provide strong evidence that our SFGN improves detection accuracy, especially in small object test.https://www.tandfonline.com/doi/10.1080/21642583.2022.2062479Convolutional neural networkssmall-size object detectionspatial fine-grained featuresonline hard example mining |
| spellingShingle | Jun Hu Yongfeng Wang Shuai Cheng Jiaxin Liu Jiawen Kang Wenxing Yang SFGNet detecting objects via spatial fine-grained feature and enhanced RPN with spatial context Systems Science & Control Engineering Convolutional neural networks small-size object detection spatial fine-grained features online hard example mining |
| title | SFGNet detecting objects via spatial fine-grained feature and enhanced RPN with spatial context |
| title_full | SFGNet detecting objects via spatial fine-grained feature and enhanced RPN with spatial context |
| title_fullStr | SFGNet detecting objects via spatial fine-grained feature and enhanced RPN with spatial context |
| title_full_unstemmed | SFGNet detecting objects via spatial fine-grained feature and enhanced RPN with spatial context |
| title_short | SFGNet detecting objects via spatial fine-grained feature and enhanced RPN with spatial context |
| title_sort | sfgnet detecting objects via spatial fine grained feature and enhanced rpn with spatial context |
| topic | Convolutional neural networks small-size object detection spatial fine-grained features online hard example mining |
| url | https://www.tandfonline.com/doi/10.1080/21642583.2022.2062479 |
| work_keys_str_mv | AT junhu sfgnetdetectingobjectsviaspatialfinegrainedfeatureandenhancedrpnwithspatialcontext AT yongfengwang sfgnetdetectingobjectsviaspatialfinegrainedfeatureandenhancedrpnwithspatialcontext AT shuaicheng sfgnetdetectingobjectsviaspatialfinegrainedfeatureandenhancedrpnwithspatialcontext AT jiaxinliu sfgnetdetectingobjectsviaspatialfinegrainedfeatureandenhancedrpnwithspatialcontext AT jiawenkang sfgnetdetectingobjectsviaspatialfinegrainedfeatureandenhancedrpnwithspatialcontext AT wenxingyang sfgnetdetectingobjectsviaspatialfinegrainedfeatureandenhancedrpnwithspatialcontext |