Task-Guided Multi-Annotation Triplet Learning for Remote Sensing Representations

2026-04-04

AI Summary

This work addresses the limitation of conventional multi-task triplet loss, which relies on static weighting schemes and struggles to dynamically balance the influence of diverse annotation tasks on shared representations. To overcome this, the authors propose a task-guided multi-annotation triplet learning framework that abandons fixed weights in favor of a novel dynamic triplet selection strategy based on inter-task mutual information. This approach samples the most informative instances in a task-aware manner, thereby driving the optimization of shared representations. By integrating mutual information estimation with multi-task representation learning, the method significantly enhances both classification and regression performance on an aerial wildlife dataset, demonstrating that task-guided sampling effectively yields more discriminative shared features.

Abstract

Prior multi-task triplet loss methods relied on static weights to balance supervision between various types of annotation. However, static weighting requires tuning and does not account for how tasks interact when shaping a shared representation. To address this, the proposed task-guided multi-annotation triplet loss removes this dependency by selecting triplets through a mutual-information criteria that identifies triplets most informative across tasks. This strategy modifies which samples influence the representation rather than adjusting loss magnitudes. Experiments on an aerial wildlife dataset compare the proposed task-guided selection against several triplet loss setups for shaping a representation in an effective multi-task manner. The results show improved classification and regression performance and demonstrate that task-aware triplet selection produces a more effective shared representation for downstream tasks.

Problem

• multi-task learning
• triplet loss
• remote sensing
• representation learning
• annotation balancing

Innovation

• task-guided triplet learning
• multi-annotation
• mutual information
• shared representation
• remote sensing