Class-agnostic Object Detection

TL;DR

This paper introduces class-agnostic object detection, focusing on detecting all objects regardless of class, and proposes new training protocols, baseline methods, and an adversarial learning framework to improve performance.

Contribution

It defines the new problem of class-agnostic detection, proposes evaluation protocols, and introduces an adversarial learning approach to enhance detection without class-specific information.

Findings

Adversarial learning improves detection accuracy.

Baseline methods provide a starting point for future research.

Evaluation protocols enable benchmarking in this new domain.

Abstract

Object detection models perform well at localizing and classifying objects that they are shown during training. However, due to the difficulty and cost associated with creating and annotating detection datasets, trained models detect a limited number of object types with unknown objects treated as background content. This hinders the adoption of conventional detectors in real-world applications like large-scale object matching, visual grounding, visual relation prediction, obstacle detection (where it is more important to determine the presence and location of objects than to find specific types), etc. We propose class-agnostic object detection as a new problem that focuses on detecting objects irrespective of their object-classes. Specifically, the goal is to predict bounding boxes for all objects in an image but not their object-classes. The predicted boxes can then be consumed by another system to perform application-specific classification, retrieval, etc. We propose training and evaluation protocols for benchmarking class-agnostic detectors to advance future research in this domain. Finally, we propose (1) baseline methods and (2) a new adversarial learning framework for class-agnostic detection that forces the model to exclude class-specific information from features used for predictions. Experimental results show that adversarial learning improves class-agnostic detection efficacy.