Fast OWDETR: transformer for open world object detection
Object detection is one of the basic computer vision tasks. Recently, a more challenging task, called open world object detection, which aims to identify novel unknown objects and incrementally learn to classify them when labels are available has been proposed. Open World Object Detector (ORE) an...
| Autor principal: | |
|---|---|
| Otros Autores: | |
| Formato: | Thesis-Master by Coursework |
| Lenguaje: | English |
| Publicado: |
Nanyang Technological University
2022
|
| Materias: | |
| Acceso en línea: | https://hdl.handle.net/10356/162462 |
| _version_ | 1826117334484385792 |
|---|---|
| author | Chen, Xuanying |
| author2 | Tan Yap Peng |
| author_facet | Tan Yap Peng Chen, Xuanying |
| author_sort | Chen, Xuanying |
| collection | NTU |
| description | Object detection is one of the basic computer vision tasks. Recently, a more
challenging task, called open world object detection, which aims to identify
novel unknown objects and incrementally learn to classify them when labels
are available has been proposed. Open World Object Detector (ORE) and
Open-world Detection Transformer (OWDETR) are two methodologies proposed
to address the open world task, while they are both time-consuming
in training and with shortcomings. Aiming to improve the training speed
and detection performance, we propose Fast OWDETR based on OWDETR
which is a transformer-based approach. Speci cally, we replace the attentiondriven
pseudo labeling mechanism in OWDETR with a logits-based one, and
change the standard Deformable DETR into Deformable DETR with box
re nement. For shorter transferring time between tasks, we present an incremental
learning approach which dynamically reduces the number of trainable
parameters in the classi cation head while keeping the backbone frozen after
initial training. Our extensive experiments show that Fast OWDETR can
achieve detection performance comparable with OWDETR while using less
training time within tasks and between tasks. |
| first_indexed | 2024-10-01T04:25:55Z |
| format | Thesis-Master by Coursework |
| id | ntu-10356/162462 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T04:25:55Z |
| publishDate | 2022 |
| publisher | Nanyang Technological University |
| record_format | dspace |
| spelling | ntu-10356/1624622023-07-04T17:50:11Z Fast OWDETR: transformer for open world object detection Chen, Xuanying Tan Yap Peng School of Electrical and Electronic Engineering EYPTan@ntu.edu.sg Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision Object detection is one of the basic computer vision tasks. Recently, a more challenging task, called open world object detection, which aims to identify novel unknown objects and incrementally learn to classify them when labels are available has been proposed. Open World Object Detector (ORE) and Open-world Detection Transformer (OWDETR) are two methodologies proposed to address the open world task, while they are both time-consuming in training and with shortcomings. Aiming to improve the training speed and detection performance, we propose Fast OWDETR based on OWDETR which is a transformer-based approach. Speci cally, we replace the attentiondriven pseudo labeling mechanism in OWDETR with a logits-based one, and change the standard Deformable DETR into Deformable DETR with box re nement. For shorter transferring time between tasks, we present an incremental learning approach which dynamically reduces the number of trainable parameters in the classi cation head while keeping the backbone frozen after initial training. Our extensive experiments show that Fast OWDETR can achieve detection performance comparable with OWDETR while using less training time within tasks and between tasks. Master of Science (Computer Control and Automation) 2022-10-21T04:45:27Z 2022-10-21T04:45:27Z 2022 Thesis-Master by Coursework Chen, X. (2022). Fast OWDETR: transformer for open world object detection. Master's thesis, Nanyang Technological University, Singapore. https://hdl.handle.net/10356/162462 https://hdl.handle.net/10356/162462 en application/pdf Nanyang Technological University |
| spellingShingle | Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision Chen, Xuanying Fast OWDETR: transformer for open world object detection |
| title | Fast OWDETR: transformer for open world object detection |
| title_full | Fast OWDETR: transformer for open world object detection |
| title_fullStr | Fast OWDETR: transformer for open world object detection |
| title_full_unstemmed | Fast OWDETR: transformer for open world object detection |
| title_short | Fast OWDETR: transformer for open world object detection |
| title_sort | fast owdetr transformer for open world object detection |
| topic | Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision |
| url | https://hdl.handle.net/10356/162462 |
| work_keys_str_mv | AT chenxuanying fastowdetrtransformerforopenworldobjectdetection |