MonoTher-Depth: Enhancing Thermal Depth Estimation via Confidence-Aware Distillation

2025-03-01

IEEE Robotics and Automation Letters

AI Summary

Monocular depth estimation (MDE) from thermal imagery suffers from poor generalization under adverse conditions—such as fog, smoke, and low-light—due to severe scarcity of annotated depth data. Method: This paper proposes a label-free cross-modal knowledge distillation framework that transfers knowledge from a pre-trained RGB-based MDE model to a thermal MDE model. Its core innovation is a confidence-aware distillation mechanism: it models the prediction confidence of the RGB model on thermal inputs and dynamically weights its outputs to mitigate domain shift. The method integrates monocular thermal-depth estimation, RGB-to-thermal cross-modal distillation, and adaptive confidence-based weighting. Results: Evaluated on unseen, unlabeled thermal scenes, the approach reduces absolute relative error by 22.88% compared to baseline methods, significantly improving both thermal depth estimation accuracy and cross-scene generalization capability.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Calibration & Uncertainty Quantification
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Monocular depth estimation (MDE) from thermal images is a crucial technology for robotic systems operating in challenging conditions such as fog, smoke, and low light. The limited availability of labeled thermal data constrains the generalization capabilities of thermal MDE models compared to foundational RGB MDE models, which benefit from datasets of millions of images across diverse scenarios. To address this challenge, we introduce a novel pipeline that enhances thermal MDE through knowledge distillation from a versatile RGB MDE model. Our approach features a confidence-aware distillation method that utilizes the predicted confidence of the RGB MDE to selectively strengthen the thermal MDE model, capitalizing on the strengths of the RGB model while mitigating its weaknesses. Our method significantly improves the accuracy of the thermal MDE, independent of the availability of labeled depth supervision, and greatly expands its applicability to new scenarios. In our experiments on new scenarios without labeled depth, the proposed confidence-aware distillation method reduces the absolute relative error of thermal MDE by 22.88% compared to the baseline without distillation.

Problem

• Enhancing thermal depth estimation with limited labeled data
• Improving generalization via RGB model knowledge distillation
• Reducing depth estimation error in challenging conditions

Innovation

• Confidence-aware distillation from RGB model
• Selective strengthening using RGB confidence
• Improves thermal depth without labeled data