DCEF<sup>2</sup>-YOLO: Aerial Detection YOLO with Deformable Convolution–Efficient Feature Fusion for Small Target Detection
Deep learning technology for real-time small object detection in aerial images can be used in various industrial environments such as real-time traffic surveillance and military reconnaissance. However, detecting small objects with few pixels and low resolution remains a challenging problem that req...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2024-03-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/16/6/1071 |
| _version_ | 1797239445114585088 |
|---|---|
| author | Yeonha Shin Heesub Shin Jaewoo Ok Minyoung Back Jaehyuk Youn Sungho Kim |
| author_facet | Yeonha Shin Heesub Shin Jaewoo Ok Minyoung Back Jaehyuk Youn Sungho Kim |
| author_sort | Yeonha Shin |
| collection | DOAJ |
| description | Deep learning technology for real-time small object detection in aerial images can be used in various industrial environments such as real-time traffic surveillance and military reconnaissance. However, detecting small objects with few pixels and low resolution remains a challenging problem that requires performance improvement. To improve the performance of small object detection, we propose DCEF<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><msup><mrow><mspace width="-2.pt"></mspace><mo> </mo></mrow><mn>2</mn></msup></semantics></math></inline-formula>-YOLO. Our proposed method enables efficient real-time small object detection by using a deformable convolution (DFConv) module and an efficient feature fusion structure to maximize the use of the internal feature information of objects. DFConv preserves small object information by preventing the mixing of object information with the background. The optimized feature fusion structure produces high-quality feature maps for efficient real-time small object detection while maximizing the use of limited information. Additionally, modifying the input data processing stage and reducing the detection layer to suit small object detection also contributes to performance improvement. When compared to the performance of the latest YOLO-based models (such as DCN-YOLO and YOLOv7), DCEF<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><msup><mrow><mspace width="-2.pt"></mspace><mo> </mo></mrow><mn>2</mn></msup></semantics></math></inline-formula>-YOLO outperforms them, with a mAP of +6.1% on the DOTA-v1.0 test set, +0.3% on the NWPU VHR-10 test set, and +1.5% on the VEDAI512 test set. Furthermore, it has a fast processing speed of 120.48 FPS with an RTX3090 for 512 × 512 images, making it suitable for real-time small object detection tasks. |
| first_indexed | 2024-04-24T17:51:39Z |
| format | Article |
| id | doaj.art-5d61a00dae5b4b9e9507962b9dbc44bb |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-04-24T17:51:39Z |
| publishDate | 2024-03-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-5d61a00dae5b4b9e9507962b9dbc44bb2024-03-27T14:02:47ZengMDPI AGRemote Sensing2072-42922024-03-01166107110.3390/rs16061071DCEF<sup>2</sup>-YOLO: Aerial Detection YOLO with Deformable Convolution–Efficient Feature Fusion for Small Target DetectionYeonha Shin0Heesub Shin1Jaewoo Ok2Minyoung Back3Jaehyuk Youn4Sungho Kim5Advanced Visual Intelligence Laboratory, Department of Electronic Engineering, Yeungnam University, 280 Daehak-ro, Gyeongsan 38541, Republic of KoreaLIG Nex1 Co., Ltd., Yongin 16911, Republic of KoreaLIG Nex1 Co., Ltd., Yongin 16911, Republic of KoreaLIG Nex1 Co., Ltd., Yongin 16911, Republic of KoreaLIG Nex1 Co., Ltd., Yongin 16911, Republic of KoreaAdvanced Visual Intelligence Laboratory, Department of Electronic Engineering, Yeungnam University, 280 Daehak-ro, Gyeongsan 38541, Republic of KoreaDeep learning technology for real-time small object detection in aerial images can be used in various industrial environments such as real-time traffic surveillance and military reconnaissance. However, detecting small objects with few pixels and low resolution remains a challenging problem that requires performance improvement. To improve the performance of small object detection, we propose DCEF<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><msup><mrow><mspace width="-2.pt"></mspace><mo> </mo></mrow><mn>2</mn></msup></semantics></math></inline-formula>-YOLO. Our proposed method enables efficient real-time small object detection by using a deformable convolution (DFConv) module and an efficient feature fusion structure to maximize the use of the internal feature information of objects. DFConv preserves small object information by preventing the mixing of object information with the background. The optimized feature fusion structure produces high-quality feature maps for efficient real-time small object detection while maximizing the use of limited information. Additionally, modifying the input data processing stage and reducing the detection layer to suit small object detection also contributes to performance improvement. When compared to the performance of the latest YOLO-based models (such as DCN-YOLO and YOLOv7), DCEF<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><msup><mrow><mspace width="-2.pt"></mspace><mo> </mo></mrow><mn>2</mn></msup></semantics></math></inline-formula>-YOLO outperforms them, with a mAP of +6.1% on the DOTA-v1.0 test set, +0.3% on the NWPU VHR-10 test set, and +1.5% on the VEDAI512 test set. Furthermore, it has a fast processing speed of 120.48 FPS with an RTX3090 for 512 × 512 images, making it suitable for real-time small object detection tasks.https://www.mdpi.com/2072-4292/16/6/1071aerial object detectionsmall target detectionreal-time object detectionDCNdeformable convolutionfeature fusion |
| spellingShingle | Yeonha Shin Heesub Shin Jaewoo Ok Minyoung Back Jaehyuk Youn Sungho Kim DCEF<sup>2</sup>-YOLO: Aerial Detection YOLO with Deformable Convolution–Efficient Feature Fusion for Small Target Detection Remote Sensing aerial object detection small target detection real-time object detection DCN deformable convolution feature fusion |
| title | DCEF<sup>2</sup>-YOLO: Aerial Detection YOLO with Deformable Convolution–Efficient Feature Fusion for Small Target Detection |
| title_full | DCEF<sup>2</sup>-YOLO: Aerial Detection YOLO with Deformable Convolution–Efficient Feature Fusion for Small Target Detection |
| title_fullStr | DCEF<sup>2</sup>-YOLO: Aerial Detection YOLO with Deformable Convolution–Efficient Feature Fusion for Small Target Detection |
| title_full_unstemmed | DCEF<sup>2</sup>-YOLO: Aerial Detection YOLO with Deformable Convolution–Efficient Feature Fusion for Small Target Detection |
| title_short | DCEF<sup>2</sup>-YOLO: Aerial Detection YOLO with Deformable Convolution–Efficient Feature Fusion for Small Target Detection |
| title_sort | dcef sup 2 sup yolo aerial detection yolo with deformable convolution efficient feature fusion for small target detection |
| topic | aerial object detection small target detection real-time object detection DCN deformable convolution feature fusion |
| url | https://www.mdpi.com/2072-4292/16/6/1071 |
| work_keys_str_mv | AT yeonhashin dcefsup2supyoloaerialdetectionyolowithdeformableconvolutionefficientfeaturefusionforsmalltargetdetection AT heesubshin dcefsup2supyoloaerialdetectionyolowithdeformableconvolutionefficientfeaturefusionforsmalltargetdetection AT jaewoook dcefsup2supyoloaerialdetectionyolowithdeformableconvolutionefficientfeaturefusionforsmalltargetdetection AT minyoungback dcefsup2supyoloaerialdetectionyolowithdeformableconvolutionefficientfeaturefusionforsmalltargetdetection AT jaehyukyoun dcefsup2supyoloaerialdetectionyolowithdeformableconvolutionefficientfeaturefusionforsmalltargetdetection AT sunghokim dcefsup2supyoloaerialdetectionyolowithdeformableconvolutionefficientfeaturefusionforsmalltargetdetection |