Abstract:Aiming at recognizing and localizing the object of novel categories by a few reference samples, few-shot object detection is a quite challenging task. Previous works often depend on the fine-tuning process to transfer their model to the novel category and rarely consider the defect of fine-tuning, resulting in many drawbacks. For example, these methods are far from satisfying in the low-shot or episode-based scenarios since the fine-tuning process in object detection requires much time and high-shot support data. To this end, this paper proposes a plug-and-play few-shot object detection (PnP-FSOD) framework that can accurately and directly detect the objects of novel categories without the fine-tuning process. To accomplish the objective, the PnP-FSOD framework contains two parallel techniques to address the core challenges in the few-shot learning, i.e., across-category task and few-annotation support. Concretely, we first propose two simple but effective meta strategies for the box classifier and RPN module to enable the across-category object detection without fine-tuning. Then, we introduce two explicit inferences into the localization process to reduce its dependence on the annotated data, including explicit localization score and semi-explicit box regression. In addition to the PnP-FSOD framework, we propose a novel one-step tuning method that can avoid the defects in fine-tuning. It is noteworthy that the proposed techniques and tuning method are based on the general object detector without other prior methods, so they are easily compatible with the existing FSOD methods. Extensive experiments show that the PnP-FSOD framework has achieved the state-of-the-art few-shot object detection performance without any tuning method. After applying the one-step tuning method, it further shows a significant lead in both efficiency, precision, and recall, under varied few-shot evaluation protocols.

Towards Stabilized Few-Shot Object Detection with Less Forgetting via Sample Normalization

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

Towards Generalized and Incremental Few-Shot Object Detection

Delve into Cosine: Few-shot Object Detection Via Adaptive Norm-cutting Scale

Fine-Grained Prototypes Distillation for Few-Shot Object Detection

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

Bilaterally Normalized Scale-Consistent Sinkhorn Distance for Few-Shot Image Classification

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

Understanding Negative Proposals in Generic Few-Shot Object Detection

Decoupled Metric Network for Single-Stage Few-Shot Object Detection

Towards Discriminative and Transferable One-Stage Few-Shot Object Detectors

Balancing Feature Alignment and Uniformity for Few-Shot Classification.

Few-Shot Object Detection Based on Adaptive Attention Mechanism and Large-Margin Softmax

PNPDet: Efficient Few-shot Detection Without Forgetting Via Plug-and-Play Sub-networks

Class Regularization: Improve Few-shot Image Classification by Reducing Meta Shift

Few-Shot Object Detection by Second-Order Pooling

MPF-Net: multi-projection filtering network for few-shot object detection

Universal-Prototype Enhancing for Few-Shot Object Detection.

Proposal Distribution Calibration for Few-Shot Object Detection

Cross attention redistribution with contrastive learning for few shot object detection

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