OW-Mamba: Mamba for open world object detection

Object detection is a fundamental task in computer vision, and recently, a more challenging variant known as open-world object detection has gained attention. This task involves not only identifying novel, unknown objects but also incrementally learning to classify...

Full description

Saved in:
Bibliographic Details
Main Author: Sun, Heyuan
Other Authors: Yap Kim Hui
Format: Thesis-Master by Coursework
Language:English
Published: Nanyang Technological University 2024
Subjects:
Online Access:https://hdl.handle.net/10356/181673
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
id sg-ntu-dr.10356-181673
record_format dspace
spelling sg-ntu-dr.10356-1816732024-12-13T15:47:37Z OW-Mamba: Mamba for open world object detection Sun, Heyuan Yap Kim Hui School of Electrical and Electronic Engineering EKHYap@ntu.edu.sg Engineering Open world object detection Object detection is a fundamental task in computer vision, and recently, a more challenging variant known as open-world object detection has gained attention. This task involves not only identifying novel, unknown objects but also incrementally learning to classify them as labels become available. Two notable approaches, Open World Detection Transformer (OWDETR) and Localization and Identification Cascade Detection Transformer (CAT), have been proposed to address this challenge. However, these methods are prone to generating false unknown objects and are computationally expensive, especially with high-resolution images. Additionally, there is significant room for improvement in detecting novel objects. To overcome these limitations, we propose OW-Mamba, an enhanced approach based on CAT. Specifically, we replace the ResNet-50 backbone in CAT with VMamba-T and introduce a dual-stream decoder, which improves both localization and classification. Furthermore, we refine the pseudo-labeling process to reduce the generation of false positives. Extensive experiments show that OW- Mamba outperforms CAT in Tasks 1, 3, and 4, while also significantly reducing the time and GPU memory required. Master's degree 2024-12-12T23:01:25Z 2024-12-12T23:01:25Z 2024 Thesis-Master by Coursework Sun, H. (2024). OW-Mamba: Mamba for open world object detection. Master's thesis, Nanyang Technological University, Singapore. https://hdl.handle.net/10356/181673 https://hdl.handle.net/10356/181673 en application/pdf Nanyang Technological University
institution Nanyang Technological University
building NTU Library
continent Asia
country Singapore
Singapore
content_provider NTU Library
collection DR-NTU
language English
topic Engineering
Open world object detection
spellingShingle Engineering
Open world object detection
Sun, Heyuan
OW-Mamba: Mamba for open world object detection
description Object detection is a fundamental task in computer vision, and recently, a more challenging variant known as open-world object detection has gained attention. This task involves not only identifying novel, unknown objects but also incrementally learning to classify them as labels become available. Two notable approaches, Open World Detection Transformer (OWDETR) and Localization and Identification Cascade Detection Transformer (CAT), have been proposed to address this challenge. However, these methods are prone to generating false unknown objects and are computationally expensive, especially with high-resolution images. Additionally, there is significant room for improvement in detecting novel objects. To overcome these limitations, we propose OW-Mamba, an enhanced approach based on CAT. Specifically, we replace the ResNet-50 backbone in CAT with VMamba-T and introduce a dual-stream decoder, which improves both localization and classification. Furthermore, we refine the pseudo-labeling process to reduce the generation of false positives. Extensive experiments show that OW- Mamba outperforms CAT in Tasks 1, 3, and 4, while also significantly reducing the time and GPU memory required.
author2 Yap Kim Hui
author_facet Yap Kim Hui
Sun, Heyuan
format Thesis-Master by Coursework
author Sun, Heyuan
author_sort Sun, Heyuan
title OW-Mamba: Mamba for open world object detection
title_short OW-Mamba: Mamba for open world object detection
title_full OW-Mamba: Mamba for open world object detection
title_fullStr OW-Mamba: Mamba for open world object detection
title_full_unstemmed OW-Mamba: Mamba for open world object detection
title_sort ow-mamba: mamba for open world object detection
publisher Nanyang Technological University
publishDate 2024
url https://hdl.handle.net/10356/181673
_version_ 1819112986778796032