The Pursuit of Knowledge: Discovering and Localizing Novel Categories using Dual Memory

TL;DR

This paper introduces a dual memory approach for discovering and localizing novel object categories in large, cluttered datasets like COCO, leveraging prior knowledge to improve detection of unknown objects.

Contribution

The paper proposes a novel dual memory framework that uses prior knowledge and memory modules to enhance object category discovery in complex, real-world datasets.

Findings

Effective discovery of novel categories in COCO dataset

Improved localization accuracy for unknown objects

Demonstrated in-the-wild applicability of the method

Abstract

We tackle object category discovery, which is the problem of discovering and localizing novel objects in a large unlabeled dataset. While existing methods show results on datasets with less cluttered scenes and fewer object instances per image, we present our results on the challenging COCO dataset. Moreover, we argue that, rather than discovering new categories from scratch, discovery algorithms can benefit from identifying what is already known and focusing their attention on the unknown. We propose a method that exploits prior knowledge about certain object types to discover new categories by leveraging two memory modules, namely Working and Semantic memory. We show the performance of our detector on the COCO minival dataset to demonstrate its in-the-wild capabilities.