Towards Accurate Localization by Instance Search

TL;DR

This paper introduces a self-paced learning framework that improves object localization accuracy by leveraging instance search results, enabling detection of unknown categories without requiring object annotations.

Contribution

It proposes a novel self-paced learning approach that localizes objects from search results, addressing unknown categories and few-shot detection within a unified framework.

Findings

Outperforms state-of-the-art methods in localization accuracy

Effective in localizing objects without category knowledge

Addresses few-shot detection with superior results

Abstract

Visual object localization is the key step in a series of object detection tasks. In the literature, high localization accuracy is achieved with the mainstream strongly supervised frameworks. However, such methods require object-level annotations and are unable to detect objects of unknown categories. Weakly supervised methods face similar difficulties. In this paper, a self-paced learning framework is proposed to achieve accurate object localization on the rank list returned by instance search. The proposed framework mines the target instance gradually from the queries and their corresponding top-ranked search results. Since a common instance is shared between the query and the images in the rank list, the target visual instance can be accurately localized even without knowing what the object category is. In addition to performing localization on instance search, the issue of few-shot object detection is also addressed under the same framework. Superior performance over state-of-the-art methods is observed on both tasks.