Cross-Domain Dead Tree Detection via Knowledge Distillation in Aerial Imagery

2026-06-01

AI Summary

This study addresses the limited generalization of dead tree detection models caused by domain shifts and scarce labeled data. To tackle this, the authors propose a feature-level knowledge distillation approach that transfers the TreeMort-1T-UNet model—originally trained on Finnish aerial imagery—to diverse forest domains in Poland, Germany, and Estonia. Integrated within a U-Net architecture, the method significantly reduces false positives and enhances domain-invariant representations under extremely limited annotation budgets, outperforming both fine-tuning and alternative distillation strategies. Experimental results demonstrate strong cross-domain performance: on the Polish dataset, the approach achieves a Mean Tree IoU of 0.106, an Instance F1-score of 0.63, Precision of 0.55, Centroid Error of 3.039, and a linear probing AUC of 0.95, confirming its robustness and effectiveness across heterogeneous forest environments.

Abstract

Detecting dead trees in aerial imagery is vital for assessing forest health, especially as tree mortality increases globally due to climate change, but domain variability and scarce labeled data often limit model generalization. This study advances the TreeMort-1T-UNet (Tree Mortality 1-Task U-Net) model, initially trained on Finnish aerial imagery (source domain), by applying knowledge distillation (KD) to adapt it to various target domains, including Polish, German, and Estonian datasets representing diverse forest types. We assess four KD variants: Basic, Self, Feature-level, and Ensemble, against a fine-tuning baseline, using Mean Tree IoU, Instance F1-score, Instance Precision, and Mean Centroid Error as key metrics, alongside representational analyses (e.g., cosine similarity, CKA, SSIM, t-SNE, and linear probing) for domain invariance. Feature-level KD outperforms others, yielding a Mean Tree IoU of 0.106, Instance F1-score of 0.63, Instance Precision of 0.55, and Mean Centroid Error of 3.039 on the Polish dataset, with robust precision across other target domains (e.g., 0.15 on Finnish, 0.67 on Polish, 0.60 on German, 0.59 on Estonian). It excels in low-data scenarios with fewer false positives and shows superior representational invariance (e.g., higher deep-layer CKA/SSIM, better domain mixing in t-SNE, and linear probing AUC of 0.95), making it ideal for precision-critical forestry applications. Additional ablation studies confirm that key components like feature alignment enhance its performance balance across metrics. Our findings demonstrate KD's potential to enhance transfer learning in remote sensing, offering a scalable, domain-robust tool for ecological monitoring and sustainable forest management.

Problem

• dead tree detection
• cross-domain
• aerial imagery
• domain variability
• limited labeled data

Innovation

• Knowledge Distillation
• Cross-Domain Adaptation
• Feature-level Distillation
• Remote Sensing
• Domain Invariance