AirDet: Few-Shot Detection Without Fine-Tuning for Autonomous Exploration

Li, Bowen; Wang, Chen; Reddy, P. Santosh; S, Kim; Scherer, Sebastian

doi:10.1007/978-3-031-19842-7_25

Cited by 19 publications

(3 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Nowadays, most of the research is carried out based on classification [22,23]. Few-shot object detection (FSOD) is a core problem in few-shot learning, which involves classification and location tasks that are more difficult to perform than classification [24,25]. FSOD research is mainly based on meta-learning and fine-tuning methods.…”

Section: Few-shot Object Detectionmentioning

confidence: 99%

Few-Shot Object Detection with Memory Contrastive Proposal Based on Semantic Priors

Xiao,

Xu,

Xiao

et al. 2023

Electronics

View full text Add to dashboard Cite

Few-shot object detection (FSOD) aims to detect objects belonging to novel classes with few training samples. With the small number of novel class samples, the visual information extracted is insufficient to accurately represent the object itself, presenting significant intra-class variance and confusion between classes of similar samples, resulting in large errors in the detection results of the novel class samples. We propose a few-shot object detection framework to achieve effective classification and detection by embedding semantic information and contrastive learning. Firstly, we introduced a semantic fusion (SF) module, which projects semantic spatial information into visual space for interaction, to compensate for the lack of visual information and further enhance the representation of feature information. To further improve the classification performance, we embed the memory contrastive proposal (MCP) module to adjust the distribution of the feature space by calculating the contrastive loss between the class-centered features of previous samples and the current input features to obtain a more discriminative embedding space for better intra-class aggregation and inter-class separation for subsequent classification and detection. Extensive experiments on the PASCAL VOC and MS-COCO datasets show that the performance of our proposed method is effectively improved. Our proposed method improves nAP50 over the baseline model by 4.5% and 3.5%.

show abstract

Section: Few-shot Object Detectionmentioning

confidence: 99%

Few-Shot Object Detection with Memory Contrastive Proposal Based on Semantic Priors

Xiao,

Xu,

Xiao

et al. 2023

Electronics

View full text Add to dashboard Cite

show abstract

“…General Object Detection refers to finding the object we need from the image and giving an accurate mark frame and category (Li et al, 2022a). Current deep learning-based object detection can be divided into two architectures: one-stage and two-stage methods.…”

Section: Related Workmentioning

confidence: 99%

“…This method can be particularly helpful for new dataset tasks if the large-scale dataset contains similar target categories (Zhu et al, 2021a) (e.g., the experience of motorbike detection can help bicycle detection in another task). However, there are domain shifts between different datasets due to differences in shots, environments, and objects themselves (Li et al, 2022a;Yu et al, 2022). These domain shifts prevent us from fully exploiting prior knowledge on large datasets.…”

Section: Introductionmentioning

confidence: 99%

See you somewhere in the ocean: few-shot domain adaptive underwater object detection

Han

Zhai

et al. 2023

Front. Mar. Sci.

View full text Add to dashboard Cite

The current data-driven underwater object detection methods have significantly progressed. However, there are millions of marine creatures in the oceans, and collecting a corresponding database for each species for similar tasks (such as object detection)is expensive. Besides, marine environments are more complex than in-air cases. Water quality, illuminations, and seafloor topography may lead to domain shifting with visual instability features of underwater objects. To tackle these problems, we propose a few-shot adaptive object detection framework with a novel two-stage training approach and a lightweight feature correction module to accommodate both image-level and instance-level domain shifting on multiple datasets. Our method can be trained in a source domain and quickly adapt to an unfamiliar target domain with only a few labeled samples. Extensive experimental results have demonstrated the knowledge transfer capability of the proposed method in detecting two similar marine species. The code will be available at: https://github.com/roadhan/FSCW

show abstract

Novel-Registrable Weights and Region-Level Contrastive Learning for Incremental Few-shot Object Detection

Tang,

Mei,

Qiu

et al. 2023

Communications in Computer and Information Science

View full text Add to dashboard Cite

AirDet: Few-Shot Detection Without Fine-Tuning for Autonomous Exploration

Cited by 19 publications

References 38 publications

Few-Shot Object Detection with Memory Contrastive Proposal Based on Semantic Priors

Few-Shot Object Detection with Memory Contrastive Proposal Based on Semantic Priors

See you somewhere in the ocean: few-shot domain adaptive underwater object detection

Novel-Registrable Weights and Region-Level Contrastive Learning for Incremental Few-shot Object Detection

Contact Info

Product

Resources

About