An Enhanced End-to-End Object Detector for Drone Aerial Imagery
DETR-like detectors have gained increasing popularity in current practical applications. However, we observe that their pipeline still suffer from several challenges, including unbalanced distribution of positive and negative samples, low-quality initial prediction boxes, and unreasonable gradient s...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10851266/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832576759056302080 |
|---|---|
| author | Quan Yu Qiang Tong Lin Miao Lin Qi Xiulei Liu |
| author_facet | Quan Yu Qiang Tong Lin Miao Lin Qi Xiulei Liu |
| author_sort | Quan Yu |
| collection | DOAJ |
| description | DETR-like detectors have gained increasing popularity in current practical applications. However, we observe that their pipeline still suffer from several challenges, including unbalanced distribution of positive and negative samples, low-quality initial prediction boxes, and unreasonable gradient structure in the decoding stage. These challenges hinder both the convergence speed and detection performance of the model. To address these issues, we propose an enhanced DETR-like model called EM-DETR. It combines three innovative methods, including Dynamic Groups Assignment, Mixed Query Re-Selection, and Look Forward Stage. Dynamic Groups Assignment employs adaptive parameters to balance the number of positive and negative samples, providing more effective supervision signals for ground-truth boxes. Mixed Query Re-Selection utilizes high-quality bounding boxes regressed by subnet to initialize decoder queries, offering superior prior information for the decoder. Look Forward Stage introduces a more rational gradient structure which eliminates inter-layer information bias between decoders. We conduct extensive experiments to evaluate the effectiveness of our proposed method. On VisDrone2021-DET, EM-DETR with ResNet50 achieved 23.9% AP after 12 epochs of training. Compared to the baseline, this represents an improvement of 4.7% AP. Moreover, the excellent performance of EM-DETR on AI-TOD and Crowdhuman proves the generalization capability of the proposed method. |
| format | Article |
| id | doaj-art-5f8503bac56d4568bcab8f3a709f291e |
| institution | Kabale University |
| issn | 2169-3536 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj-art-5f8503bac56d4568bcab8f3a709f291e2025-01-31T00:00:54ZengIEEEIEEE Access2169-35362025-01-0113187981881310.1109/ACCESS.2025.353303710851266An Enhanced End-to-End Object Detector for Drone Aerial ImageryQuan Yu0https://orcid.org/0009-0007-2635-7322Qiang Tong1https://orcid.org/0000-0003-0508-0727Lin Miao2Lin Qi3Xiulei Liu4https://orcid.org/0000-0002-9303-3682Beijing Advanced Innovation Center for Materials Genome Engineering, Beijing Information Science and Technology University, Beijing, ChinaBeijing Advanced Innovation Center for Materials Genome Engineering, Beijing Information Science and Technology University, Beijing, ChinaBeijing Advanced Innovation Center for Materials Genome Engineering, Beijing Information Science and Technology University, Beijing, ChinaSchool of Economics and Management, Beijing Information Science and Technology University, Beijing, ChinaBeijing Advanced Innovation Center for Materials Genome Engineering, Beijing Information Science and Technology University, Beijing, ChinaDETR-like detectors have gained increasing popularity in current practical applications. However, we observe that their pipeline still suffer from several challenges, including unbalanced distribution of positive and negative samples, low-quality initial prediction boxes, and unreasonable gradient structure in the decoding stage. These challenges hinder both the convergence speed and detection performance of the model. To address these issues, we propose an enhanced DETR-like model called EM-DETR. It combines three innovative methods, including Dynamic Groups Assignment, Mixed Query Re-Selection, and Look Forward Stage. Dynamic Groups Assignment employs adaptive parameters to balance the number of positive and negative samples, providing more effective supervision signals for ground-truth boxes. Mixed Query Re-Selection utilizes high-quality bounding boxes regressed by subnet to initialize decoder queries, offering superior prior information for the decoder. Look Forward Stage introduces a more rational gradient structure which eliminates inter-layer information bias between decoders. We conduct extensive experiments to evaluate the effectiveness of our proposed method. On VisDrone2021-DET, EM-DETR with ResNet50 achieved 23.9% AP after 12 epochs of training. Compared to the baseline, this represents an improvement of 4.7% AP. Moreover, the excellent performance of EM-DETR on AI-TOD and Crowdhuman proves the generalization capability of the proposed method.https://ieeexplore.ieee.org/document/10851266/Drone aerial imageryobject detectionend-to-end object detectordetection transformer |
| spellingShingle | Quan Yu Qiang Tong Lin Miao Lin Qi Xiulei Liu An Enhanced End-to-End Object Detector for Drone Aerial Imagery IEEE Access Drone aerial imagery object detection end-to-end object detector detection transformer |
| title | An Enhanced End-to-End Object Detector for Drone Aerial Imagery |
| title_full | An Enhanced End-to-End Object Detector for Drone Aerial Imagery |
| title_fullStr | An Enhanced End-to-End Object Detector for Drone Aerial Imagery |
| title_full_unstemmed | An Enhanced End-to-End Object Detector for Drone Aerial Imagery |
| title_short | An Enhanced End-to-End Object Detector for Drone Aerial Imagery |
| title_sort | enhanced end to end object detector for drone aerial imagery |
| topic | Drone aerial imagery object detection end-to-end object detector detection transformer |
| url | https://ieeexplore.ieee.org/document/10851266/ |
| work_keys_str_mv | AT quanyu anenhancedendtoendobjectdetectorfordroneaerialimagery AT qiangtong anenhancedendtoendobjectdetectorfordroneaerialimagery AT linmiao anenhancedendtoendobjectdetectorfordroneaerialimagery AT linqi anenhancedendtoendobjectdetectorfordroneaerialimagery AT xiuleiliu anenhancedendtoendobjectdetectorfordroneaerialimagery AT quanyu enhancedendtoendobjectdetectorfordroneaerialimagery AT qiangtong enhancedendtoendobjectdetectorfordroneaerialimagery AT linmiao enhancedendtoendobjectdetectorfordroneaerialimagery AT linqi enhancedendtoendobjectdetectorfordroneaerialimagery AT xiuleiliu enhancedendtoendobjectdetectorfordroneaerialimagery |