Sampling-invariant fully metric learning for few-shot object detection

Leng, Jiaxu; Chen, Taiyue; Gao, Xinbo; Mo, Mengjingcheng; Yu, Yongtao; Zhang, Yan

doi:10.1016/j.neucom.2022.09.040

Cited by 5 publications

(9 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Class prototypes can be distorted due to data shifts, which can severely affect precision. Leng et al [25] argue that class prototypes can be distorted since data shift, which can severely affect precision. We believe that this situation is related to the classifier in the head of the model.…”

Section: Analysis On Catastrophic Forgettingmentioning

confidence: 99%

“…Samples that are far apart in the feature space may be close together after data shifts, thereby creating confounding and degrading average precision. Although FCSE [16] and SIF-Net [25] alleviate catastrophic forgetting by introducing an additional deep metric branch, it greatly increases the resource consumption of the model. In addition, Hou et al [29] noted that in models using a regression layer for classification, the average precision (AP) of the new classes degraded primarily because the model confused the old and new categories.…”

Section: Analysis On Catastrophic Forgettingmentioning

confidence: 99%

“…It suggested that during the fine-tuning process, the average precision (AP) of the model decreases mainly because the model confuses the old and new categories. Class prototypes can be distorted due to data shifts, which can severely affect performance [25]. To avoid this problem, we directly compare the feature map without data shifts.…”

Section: Analysis On Catastrophic Forgettingmentioning

confidence: 99%

“…Non-supporting RoIs can increase the burden of subsequent classification and even lead to misclassification. (2) Catastrophic forgetting [8,16,[20][21][22][23][24][25][26][27][28][29]. The model gains the ability to detect new classes, but the average accuracy on the base class decreases severely, which is known as catastrophic forgetting.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

FRDet: Few‐shot object detection via feature reconstruction

Chen

Mao

Qian

et al. 2023

IET Image Processing

View full text Add to dashboard Cite

State‐of‐the‐art object detection models rely on large‐scale datasets for training to achieve good precision. Without sufficient samples, the model can suffer from severe overfitting. Current explorations in few‐shot object detection are mainly divided into meta‐learning‐based methods and fine‐tuning‐based methods. However, existing models do not focus on how feature maps should be processed to present more accurate regions of interest (RoIs), leading to many non‐supporting RoIs. These non‐supporting RoIs can increase the burden of subsequent classification and even lead to misclassification. Additionally, catastrophic forgetting is inevitable in both few‐shot object detection models. Many models classify directly in low‐dimensional spaces due to insufficient resources, but this transformation of the data space can confuse some categories and lead to misclassification. To address these problems, the Feature Reconstruction Detector (FRDet) is proposed, a simple yet effective fine‐tune‐based approach for few‐shot object detection. FRDet includes a region proposal network (RPN) based on channel attention and space attention called Multi‐Attention RPN (MARPN) and a head based on feature reconstruction called Feature Reconstruction Head (FRHead). MARPN utilizes channel attention to suppress non‐supporting classes and spatial attention to enhance support classes based on Attention RPN, resulting in fewer but more accurate RoIs. Meanwhile, FRHead utilizes support features to reconstruct query RoI features through a closed‐form solution, allowing for a comprehensive and fine‐grained comparison. The model was validated on the PASCAL VOC, MS COCO, FSOD, and CUB200 datasets and achieved better results.

show abstract

Section: Analysis On Catastrophic Forgettingmentioning

confidence: 99%

Section: Analysis On Catastrophic Forgettingmentioning

confidence: 99%

Section: Analysis On Catastrophic Forgettingmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

FRDet: Few‐shot object detection via feature reconstruction

Chen

Mao

Qian

et al. 2023

IET Image Processing

View full text Add to dashboard Cite

show abstract

“…The support branch is responsible for providing task‐specific parameters, and the query branch is in charge of combining task‐specific parameters and query features. The two branches exchange information through various fusion nodes and aggregators [12]. Fine‐tuning‐based methods have only one branch, just like traditional object detection modules.…”

Section: Introductionmentioning

confidence: 99%

Few‐shot object detection via class encoding and multi‐target decoding

Guo

Yang

Wei

et al. 2023

IET Cyber-Syst and Robotics

View full text Add to dashboard Cite

The task of few-shot object detection is to classify and locate objects through a few annotated samples. Although many studies have tried to solve this problem, the results are still not satisfactory. Recent studies have found that the class margin significantly impacts the classification and representation of the targets to be detected. Most methods use the loss function to balance the class margin, but the results show that the loss-based methods only have a tiny improvement on the few-shot object detection problem. In this study, the authors propose a class encoding method based on the transformer to balance the class margin, which can make the model pay more attention to the essential information of the features, thus increasing the recognition ability of the sample. Besides, the authors propose a multi-target decoding method to aggregate RoI vectors generated from multi-target images with multiple support vectors, which can significantly improve the detection ability of the detector for multi-target images. Experiments on Pascal visual object classes (VOC) and Microsoft Common Objects in Context datasets show that our proposed Few-Shot Object Detection via Class Encoding and Multi-Target Decoding significantly improves upon baseline detectors (average accuracy improvement is up to 10.8% on VOC and 2.1% on COCO), achieving competitive performance. In general, we propose a new way to regulate the class margin between support set vectors and a way of feature aggregation for images containing multiple objects and achieve remarkable results. Our method is implemented on mmfewshot, and the code will be available later.

show abstract