Few-shot object detection on aerial imagery via deep metric learning and knowledge inheritance
Object detection is crucial in aerial imagery analysis. Previous methods based on convolutional neural networks (CNNs) require large-scale labeled datasets for training to achieve significant success. However, the acquisition and manual annotation of such data is time-consuming and expensive. In thi...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2023-08-01
|
| Series: | International Journal of Applied Earth Observations and Geoinformation |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S1569843223002212 |
| _version_ | 1797737962815881216 |
|---|---|
| author | Wu-zhou Li Jia-wei Zhou Xiang Li Yi Cao Guang Jin |
| author_facet | Wu-zhou Li Jia-wei Zhou Xiang Li Yi Cao Guang Jin |
| author_sort | Wu-zhou Li |
| collection | DOAJ |
| description | Object detection is crucial in aerial imagery analysis. Previous methods based on convolutional neural networks (CNNs) require large-scale labeled datasets for training to achieve significant success. However, the acquisition and manual annotation of such data is time-consuming and expensive. In this study, we present an original few-shot object detection (FSOD) method that focuses on detecting unseen objects in aerial imagery with limited labeled samples. Specifically, we revisited the multi-similarity network from deep metric learning and incorporated it into a faster region-CNN (R-CNN) architecture for FSOD, learning distinctive feature representations, and effectively improving the performance of unseen class samples. Furthermore, we preserved the knowledge learned from abundant base data by designing a knowledge inheritance module to ease the influence of catastrophic forgetting. We conducted experiments on two benchmark remote sensing image datasets, and the results demonstrated that the proposed methods could achieve a satisfactory performance for FSOD in aerial imagery. |
| first_indexed | 2024-03-12T13:37:02Z |
| format | Article |
| id | doaj.art-1809efcece724912a405855ee00b5130 |
| institution | Directory Open Access Journal |
| issn | 1569-8432 |
| language | English |
| last_indexed | 2024-03-12T13:37:02Z |
| publishDate | 2023-08-01 |
| publisher | Elsevier |
| record_format | Article |
| series | International Journal of Applied Earth Observations and Geoinformation |
| spelling | doaj.art-1809efcece724912a405855ee00b51302023-08-24T04:34:09ZengElsevierInternational Journal of Applied Earth Observations and Geoinformation1569-84322023-08-01122103397Few-shot object detection on aerial imagery via deep metric learning and knowledge inheritanceWu-zhou Li0Jia-wei Zhou1Xiang Li2Yi Cao3Guang Jin4School of Remote Sensing and Information Engineering, Wuhan University, Wuhan 430022, ChinaSchool of Electronic Information, Wuhan University, Wuhan 430022, ChinaElectrical and Computer Engineering, King Abdullah University of Science and Technology, Thuwal, Saudi ArabiaSchool of Electronic Information, Wuhan University, Wuhan 430022, ChinaSchool of Remote Sensing and Information Engineering, Wuhan University, Wuhan 430022, China; Corresponding author.Object detection is crucial in aerial imagery analysis. Previous methods based on convolutional neural networks (CNNs) require large-scale labeled datasets for training to achieve significant success. However, the acquisition and manual annotation of such data is time-consuming and expensive. In this study, we present an original few-shot object detection (FSOD) method that focuses on detecting unseen objects in aerial imagery with limited labeled samples. Specifically, we revisited the multi-similarity network from deep metric learning and incorporated it into a faster region-CNN (R-CNN) architecture for FSOD, learning distinctive feature representations, and effectively improving the performance of unseen class samples. Furthermore, we preserved the knowledge learned from abundant base data by designing a knowledge inheritance module to ease the influence of catastrophic forgetting. We conducted experiments on two benchmark remote sensing image datasets, and the results demonstrated that the proposed methods could achieve a satisfactory performance for FSOD in aerial imagery.http://www.sciencedirect.com/science/article/pii/S1569843223002212Few-shot object detectionAerial imageryDeep metric learningCatastrophic forgetting |
| spellingShingle | Wu-zhou Li Jia-wei Zhou Xiang Li Yi Cao Guang Jin Few-shot object detection on aerial imagery via deep metric learning and knowledge inheritance International Journal of Applied Earth Observations and Geoinformation Few-shot object detection Aerial imagery Deep metric learning Catastrophic forgetting |
| title | Few-shot object detection on aerial imagery via deep metric learning and knowledge inheritance |
| title_full | Few-shot object detection on aerial imagery via deep metric learning and knowledge inheritance |
| title_fullStr | Few-shot object detection on aerial imagery via deep metric learning and knowledge inheritance |
| title_full_unstemmed | Few-shot object detection on aerial imagery via deep metric learning and knowledge inheritance |
| title_short | Few-shot object detection on aerial imagery via deep metric learning and knowledge inheritance |
| title_sort | few shot object detection on aerial imagery via deep metric learning and knowledge inheritance |
| topic | Few-shot object detection Aerial imagery Deep metric learning Catastrophic forgetting |
| url | http://www.sciencedirect.com/science/article/pii/S1569843223002212 |
| work_keys_str_mv | AT wuzhouli fewshotobjectdetectiononaerialimageryviadeepmetriclearningandknowledgeinheritance AT jiaweizhou fewshotobjectdetectiononaerialimageryviadeepmetriclearningandknowledgeinheritance AT xiangli fewshotobjectdetectiononaerialimageryviadeepmetriclearningandknowledgeinheritance AT yicao fewshotobjectdetectiononaerialimageryviadeepmetriclearningandknowledgeinheritance AT guangjin fewshotobjectdetectiononaerialimageryviadeepmetriclearningandknowledgeinheritance |