Modular Sensor Fusion for Semantic Segmentation

TL;DR

This paper explores modular statistical sensor fusion methods for semantic segmentation in robotics, improving robustness and performance over single modalities without requiring aligned multi-sensor training data.

Contribution

It introduces modular statistical fusion approaches that enable flexible training and improve segmentation IoU by up to 5% over single sensors.

Findings

Improved IoU performance over single modalities by up to 5%.

Modular fusion approaches require less calibration data.

Effective on both real-world and simulated datasets.

Abstract

Sensor fusion is a fundamental process in robotic systems as it extends the perceptual range and increases robustness in real-world operations. Current multi-sensor deep learning based semantic segmentation approaches do not provide robustness to under-performing classes in one modality, or require a specific architecture with access to the full aligned multi-sensor training data. In this work, we analyze statistical fusion approaches for semantic segmentation that overcome these drawbacks while keeping a competitive performance. The studied approaches are modular by construction, allowing to have different training sets per modality and only a much smaller subset is needed to calibrate the statistical models. We evaluate a range of statistical fusion approaches and report their performance against state-of-the-art baselines on both real-world and simulated data. In our experiments, the approach improves performance in IoU over the best single modality segmentation results by up to 5%. We make all implementations and configurations publicly available.