Multimodal Deep Learning for Robust RGB-D Object Recognition

TL;DR

This paper introduces a novel multimodal CNN architecture for robust RGB-D object recognition, effectively handling imperfect sensor data through innovative encoding, data augmentation, and multi-stage training, achieving state-of-the-art results.

Contribution

It proposes a new RGB-D recognition architecture with a multi-stage training process and novel depth data encoding and augmentation techniques for improved robustness.

Findings

Achieved state-of-the-art results on RGB-D object dataset.

Demonstrated robustness in noisy real-world RGB-D scenarios.

Effective depth encoding enables learning without large depth datasets.

Abstract

Robust object recognition is a crucial ingredient of many, if not all, real-world robotics applications. This paper leverages recent progress on Convolutional Neural Networks (CNNs) and proposes a novel RGB-D architecture for object recognition. Our architecture is composed of two separate CNN processing streams - one for each modality - which are consecutively combined with a late fusion network. We focus on learning with imperfect sensor data, a typical problem in real-world robotics tasks. For accurate learning, we introduce a multi-stage training methodology and two crucial ingredients for handling depth data with CNNs. The first, an effective encoding of depth information for CNNs that enables learning without the need for large depth datasets. The second, a data augmentation scheme for robust learning with depth images by corrupting them with realistic noise patterns. We present state-of-the-art results on the RGB-D object dataset and show recognition in challenging RGB-D real-world noisy settings.