Abstract:Few-shot object detection (FSOD) aims at designing models that can accurately detect targets of novel classes in a scarce data regime. Existing research has improved detection performance with meta-learning-based models. However, existing methods continue to exhibit certain imperfections: (1) Only the interacting global features of query and support images lead to ignoring local critical features in the imprecise localization of objects from new categories. (2) Convolutional neural networks (CNNs) encounter difficulty in learning diverse pose features from exceedingly limited labeled samples of unseen classes. (3) Local context information is not fully utilized in a global attention mechanism, which means the attention modules need to be improved. As a result, the detection performance of novel-class objects is compromised. To overcome these challenges, a few-shot object detection network is proposed with a local feature enhancement module and an intrinsic feature transformation module. In this paper, a local feature enhancement module (LFEM) is designed to raise the importance of intrinsic features of the novel-class samples. In addition, an Intrinsic Feature Transform Module (IFTM) is explored to enhance the feature representation of novel-class samples, which enriches the feature space of novel classes. Finally, a more effective cross-attention module, called Global Cross-Attention Network (GCAN), which fully aggregates local and global context information between query and support images, is proposed in this paper. The crucial features of novel-class objects are extracted effectively by our model before the feature fusion between query images and support images. Our proposed method increases, on average, the detection performance by 0.93 (nAP) in comparison with previous models on the PASCAL VOC FSOD benchmark dataset. Extensive experiments demonstrate the effectiveness of our modules under various experimental settings.

Few-Shot Object Detection with Proposal Balance Refinement

Proposal Distribution Calibration for Few-Shot Object Detection

Few-Shot Object Detection Via Sample Processing

Few-Shot Object Detection with Local Feature Enhancement and Feature Interrelation

Understanding Negative Proposals in Generic Few-Shot Object Detection

Meta Faster R-CNN: Towards Accurate Few-Shot Object Detection with Attentive Feature Alignment

Robust Few-Shot Aerial Image Object Detection via Unbiased Proposals Filtration

Boosting Few-Shot Object Detection with Discriminative Representation and Class Margin

Fine-Grained Prototypes Distillation for Few-Shot Object Detection

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

Improved Region Proposal Network for Enhanced Few-Shot Object Detection

Semantic Enhanced Few-shot Object Detection

Mutually Reinforcing Structure with Proposal Contrastive Consistency for Few-Shot Object Detection.

Few-shot Object Detection Via Refining Eigenspace.

Exploring Effective Knowledge Transfer for Few-shot Object Detection

Gated Enhanced RPN and Hybrid-View for Few-Shot Object Detection

Few-Shot Object Detection Based on Self-Knowledge Distillation

Plug-and-Play Few-shot Object Detection with Meta Strategy and Explicit Localization Inference

Enhancing Few-Shot Object Detection with Modified Faster R-CNN and Transfer Learning

Extreme R-CNN: Few-Shot Object Detection via Sample Synthesis and Knowledge Distillation