Prototype-Driven Adaptation for Few-Shot Object Detection

TL;DR

This paper introduces Prototype-Driven Alignment (PDA), a lightweight method that enhances few-shot object detection by using prototype-based similarity measures, improving novel class detection with minimal extra computation.

Contribution

The paper proposes PDA, a novel prototype-based metric head for FSOD that reduces bias and improves performance without adding class-specific parameters.

Findings

PDA improves novel-class detection accuracy on VOC and GFSOD benchmarks.

PDA maintains base-class performance with minimal computational overhead.

Prototypes can be effectively adapted during fine-tuning without class-specific parameters.

Abstract

Few-shot object detection (FSOD) often suffers from base-class bias and unstable calibration when only a few novel samples are available. We propose Prototype-Driven Alignment (PDA), a lightweight, plug-in metric head for DeFRCN that provides a prototype-based "second opinion" complementary to the linear classifier. PDA maintains support-only prototypes in a learnable identity-initialized projection space and optionally applies prototype-conditioned RoI alignment to reduce geometric mismatch. During fine-tuning, prototypes can be adapted via exponential moving average(EMA) updates on labeled foreground RoIs-without introducing class-specific parameters-and are frozen at inference to ensure strict protocol compliance. PDA employs a best-of-K matching scheme to capture intra-class multi-modality and temperature-scaled fusion to combine metric similarities with detector logits. Experiments on VOC FSOD and GFSOD benchmarks show that PDA consistently improves novel-class performance with minimal impact on base classes and negligible computational overhead.