LIX: Implicitly Infusing Spatial Geometric Prior Knowledge into Visual Semantic Segmentation for Autonomous Driving

TL;DR

This paper proposes the LIX framework that uses knowledge distillation to implicitly incorporate spatial geometric prior knowledge into single-modal visual semantic segmentation networks, enhancing their performance in autonomous driving.

Contribution

It introduces a novel knowledge distillation approach with dynamic logit weighting and adaptive feature recalibration techniques for improved semantic segmentation.

Findings

LIX outperforms state-of-the-art methods on multiple datasets.

The dynamic weight controller improves knowledge transfer effectiveness.

Adaptive feature distillation enhances segmentation accuracy.

Abstract

Despite the impressive performance achieved by data-fusion networks with duplex encoders for visual semantic segmentation, they become ineffective when spatial geometric data are not available. Implicitly infusing the spatial geometric prior knowledge acquired by a data-fusion teacher network into a single-modal student network is a practical, albeit less explored research avenue. This article delves into this topic and resorts to knowledge distillation approaches to address this problem. We introduce the Learning to Infuse ''X'' (LIX) framework, with novel contributions in both logit distillation and feature distillation aspects. We present a mathematical proof that underscores the limitation of using a single, fixed weight in decoupled knowledge distillation and introduce a logit-wise dynamic weight controller as a solution to this issue. Furthermore, we develop an adaptively-recalibrated feature distillation algorithm, including two novel techniques: feature recalibration via kernel regression and in-depth feature consistency quantification via centered kernel alignment. Extensive experiments conducted with intermediate-fusion and late-fusion networks across various public datasets provide both quantitative and qualitative evaluations, demonstrating the superior performance of our LIX framework when compared to other state-of-the-art approaches.