Cos R-CNN for Online Few-shot Object Detection

TL;DR

Cos R-CNN introduces a cosine similarity-based approach for online few-shot object detection, enabling rapid adaptation to new classes without fine-tuning and outperforming existing methods on benchmark datasets.

Contribution

It presents a novel exemplar-based R-CNN framework utilizing cosine similarity for effective online few-shot detection without fine-tuning.

Findings

Achieves state-of-the-art results on 5-way ImageNet few-shot detection benchmark.

Outperforms existing methods by over 8% in 1-shot, 3% in 5-shot, and 1% in 10-shot scenarios.

Improves performance by up to 20% on novel classes in VOC dataset.

Abstract

We propose Cos R-CNN, a simple exemplar-based R-CNN formulation that is designed for online few-shot object detection. That is, it is able to localise and classify novel object categories in images with few examples without fine-tuning. Cos R-CNN frames detection as a learning-to-compare task: unseen classes are represented as exemplar images, and objects are detected based on their similarity to these exemplars. The cosine-based classification head allows for dynamic adaptation of classification parameters to the exemplar embedding, and encourages the clustering of similar classes in embedding space without the need for manual tuning of distance-metric hyperparameters. This simple formulation achieves best results on the recently proposed 5-way ImageNet few-shot detection benchmark, beating the online 1/5/10-shot scenarios by more than 8/3/1%, as well as performing up to 20% better in online 20-way few-shot VOC across all shots on novel classes.