Multi-Task Partially Supervised Learning for Super-Resolution and Semantic Segmentation on Earth Observation Data

2026-09-01

International Conference on Information Photonics

AI Summary

This study addresses the prohibitive cost of fully annotated data for joint super-resolution and semantic segmentation in Earth observation. To overcome this limitation, we propose a partially supervised multi-task learning paradigm that enables training with single-task annotated samples. Specifically, we introduce a novel hybrid variant network architecture and design a reprojection loss function to effectively exploit shared representations while explicitly enforcing improvements in super-resolved image quality. Experimental results demonstrate that the proposed method significantly reduces reliance on fully annotated datasets and outperforms existing state-of-the-art sequential approaches, achieving synergistic performance gains across both tasks.

Abstract

Super-resolution and semantic segmentation are known to benefit one another, especially in the Earth observation context. However, learning both tasks in a joint model often requires both task annotations, which is impractical and expensive. In this paper, we study the multi-task partially supervised learning paradigm for both tasks, where each example is assumed to have only a single-task annotation. To that end, we examine two multi-task architectural variations, the sequential and shared variants, and then propose a hybrid variant and a re-projection loss to benefit from the shared representation and enforce image quality of super-resolution when training with semantic segmentation. Experiments show favorable results compared to the SOTA sequential variant. Source code will be published at https://github.com/lhoangan/munera.

Problem

• Multi-Task Learning
• Partially Supervised Learning
• Super-Resolution
• Semantic Segmentation
• Earth Observation

Innovation

• Multi-Task Learning
• Partially Supervised Learning
• Super-Resolution
• Semantic Segmentation
• Re-projection Loss