Improving Object Detectors by Exploiting Bounding Boxes for Augmentation Design
Recent advancements in developing pre-trained models using large-scale datasets have emphasized the importance of robust protocols to adapt them effectively to domain-specific data, especially when the available data is limited. To achieve data-efficient fine-tuning of pre-trained object detection m...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2023-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10266321/ |
| _version_ | 1797663097308053504 |
|---|---|
| author | S. Devi Kowshik Thopalli R. Dayana P. Malarvezhi Jayaraman J. Thiagarajan |
| author_facet | S. Devi Kowshik Thopalli R. Dayana P. Malarvezhi Jayaraman J. Thiagarajan |
| author_sort | S. Devi |
| collection | DOAJ |
| description | Recent advancements in developing pre-trained models using large-scale datasets have emphasized the importance of robust protocols to adapt them effectively to domain-specific data, especially when the available data is limited. To achieve data-efficient fine-tuning of pre-trained object detection models, data augmentations are crucial. However, selecting the appropriate augmentation policy for a given dataset is known to be challenging. In this study, we address an overlooked aspect of this problem - can bounding box annotations be utilized to develop more effective augmentation policies? Our approach InterAug reveals that, by leveraging the annotations, one can deduce the optimal context for each object in a scene, rather than manipulating the entire scene or just the pre-defined bounding boxes. Through rigorous empirical research involving various benchmarks and architectures, we showcase the effectiveness of InterAug in enhancing robustness, handling data scarcity, and maintaining resilience to diverse high background contexts. An important advantage of InterAug is its compatibility with any off-the-shelf policy, requiring no modifications to the model architecture, and it significantly outperforms existing protocols. We will release the codes upon acceptance. Our codes can be found at <uri>https://github.com/kowshikthopalli/InterAug</uri>. |
| first_indexed | 2024-03-11T19:09:37Z |
| format | Article |
| id | doaj.art-e56af13db69242bf9b4d9585cf28ac01 |
| institution | Directory Open Access Journal |
| issn | 2169-3536 |
| language | English |
| last_indexed | 2024-03-11T19:09:37Z |
| publishDate | 2023-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj.art-e56af13db69242bf9b4d9585cf28ac012023-10-09T23:00:41ZengIEEEIEEE Access2169-35362023-01-011110835610836410.1109/ACCESS.2023.332063810266321Improving Object Detectors by Exploiting Bounding Boxes for Augmentation DesignS. Devi0Kowshik Thopalli1R. Dayana2https://orcid.org/0000-0003-4430-5765P. Malarvezhi3Jayaraman J. Thiagarajan4Department of Electronics and Communication Engineering, SRM Institute of Science and Technology, Kattankulathur, Chennai, IndiaLawrence Livermore National Laboratory, Livermore, CA, USADepartment of Electronics and Communication Engineering, SRM Institute of Science and Technology, Kattankulathur, Chennai, IndiaIHCM (Research and Development), ADP India, Chennai, IndiaLawrence Livermore National Laboratory, Livermore, CA, USARecent advancements in developing pre-trained models using large-scale datasets have emphasized the importance of robust protocols to adapt them effectively to domain-specific data, especially when the available data is limited. To achieve data-efficient fine-tuning of pre-trained object detection models, data augmentations are crucial. However, selecting the appropriate augmentation policy for a given dataset is known to be challenging. In this study, we address an overlooked aspect of this problem - can bounding box annotations be utilized to develop more effective augmentation policies? Our approach InterAug reveals that, by leveraging the annotations, one can deduce the optimal context for each object in a scene, rather than manipulating the entire scene or just the pre-defined bounding boxes. Through rigorous empirical research involving various benchmarks and architectures, we showcase the effectiveness of InterAug in enhancing robustness, handling data scarcity, and maintaining resilience to diverse high background contexts. An important advantage of InterAug is its compatibility with any off-the-shelf policy, requiring no modifications to the model architecture, and it significantly outperforms existing protocols. We will release the codes upon acceptance. Our codes can be found at <uri>https://github.com/kowshikthopalli/InterAug</uri>.https://ieeexplore.ieee.org/document/10266321/Object detectiondeep neural networksdata augmentationlimited datarobustness |
| spellingShingle | S. Devi Kowshik Thopalli R. Dayana P. Malarvezhi Jayaraman J. Thiagarajan Improving Object Detectors by Exploiting Bounding Boxes for Augmentation Design IEEE Access Object detection deep neural networks data augmentation limited data robustness |
| title | Improving Object Detectors by Exploiting Bounding Boxes for Augmentation Design |
| title_full | Improving Object Detectors by Exploiting Bounding Boxes for Augmentation Design |
| title_fullStr | Improving Object Detectors by Exploiting Bounding Boxes for Augmentation Design |
| title_full_unstemmed | Improving Object Detectors by Exploiting Bounding Boxes for Augmentation Design |
| title_short | Improving Object Detectors by Exploiting Bounding Boxes for Augmentation Design |
| title_sort | improving object detectors by exploiting bounding boxes for augmentation design |
| topic | Object detection deep neural networks data augmentation limited data robustness |
| url | https://ieeexplore.ieee.org/document/10266321/ |
| work_keys_str_mv | AT sdevi improvingobjectdetectorsbyexploitingboundingboxesforaugmentationdesign AT kowshikthopalli improvingobjectdetectorsbyexploitingboundingboxesforaugmentationdesign AT rdayana improvingobjectdetectorsbyexploitingboundingboxesforaugmentationdesign AT pmalarvezhi improvingobjectdetectorsbyexploitingboundingboxesforaugmentationdesign AT jayaramanjthiagarajan improvingobjectdetectorsbyexploitingboundingboxesforaugmentationdesign |