Adding New Categories in Object Detection Using Few-Shot Copy-Paste

TL;DR

This paper explores data-efficient methods for adding new object categories to detection models using few-shot copy-paste techniques, emphasizing occlusion-aware data augmentation to improve real-world applicability.

Contribution

It introduces a simple occlusion-based data augmentation strategy that enables effective addition of new categories with minimal data, achieving high accuracy in large-scale detection tasks.

Findings

Adding 15 images of a new category yields 95% accuracy on unseen test data.

Occlusion-aware augmentation significantly improves new category detection.

Simple occlusion mechanisms are surprisingly effective for few-shot learning.

Abstract

Developing data-efficient instance detection models that can handle rare object categories remains a key challenge in computer vision. However, existing research often overlooks data collection strategies and evaluation metrics tailored to real-world scenarios involving neural networks. In this study, we systematically investigate data collection and augmentation techniques focused on object occlusion, aiming to mimic occlusion relationships observed in practical applications. Surprisingly, we find that even a simple occlusion mechanism is sufficient to achieve strong performance when introducing new object categories. Notably, by adding just 15 images of a new category to a large-scale training dataset containing over half a million images across hundreds of categories, the model achieves 95\% accuracy on an unseen test set with thousands of instances of the new category.