LLM-Assisted Semantic Guidance for Sparsely Annotated Remote Sensing Object Detection

2025-09-21

AI Summary

To address the sparse-label detection challenge in remote sensing imagery—caused by dense object distributions and severe class imbalance—this paper proposes a large language model (LLM)-assisted semantic-guided framework. The method leverages the LLM’s semantic reasoning capability to generate high-confidence pseudo-labels, and introduces a class-aware dense pseudo-label assignment mechanism alongside an adaptive hard negative reweighting module to jointly mitigate pseudo-label ambiguity and background interference. The resulting end-to-end detector significantly enhances localization and classification robustness under sparse supervision. Evaluated on DOTA and HRSC2016, our approach substantially outperforms existing single-stage detectors: with only 10% of annotated data, it achieves 92.3% mAP, demonstrating the effectiveness and generalizability of semantic prior-driven sparse supervision learning.

Technology Category

• Computer Vision: Remote Sensing / Geospatial AI
• Machine Learning: Semi-Supervised Learning
• Natural Language Processing: (Large) Language Models

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Large language models for search
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

Sparse annotation in remote sensing object detection poses significant challenges due to dense object distributions and category imbalances. Although existing Dense Pseudo-Label methods have demonstrated substantial potential in pseudo-labeling tasks, they remain constrained by selection ambiguities and inconsistencies in confidence estimation.In this paper, we introduce an LLM-assisted semantic guidance framework tailored for sparsely annotated remote sensing object detection, exploiting the advanced semantic reasoning capabilities of large language models (LLMs) to distill high-confidence pseudo-labels.By integrating LLM-generated semantic priors, we propose a Class-Aware Dense Pseudo-Label Assignment mechanism that adaptively assigns pseudo-labels for both unlabeled and sparsely labeled data, ensuring robust supervision across varying data distributions. Additionally, we develop an Adaptive Hard-Negative Reweighting Module to stabilize the supervised learning branch by mitigating the influence of confounding background information. Extensive experiments on DOTA and HRSC2016 demonstrate that the proposed method outperforms existing single-stage detector-based frameworks, significantly improving detection performance under sparse annotations.

Problem

• Addressing sparse annotation challenges in remote sensing object detection
• Overcoming selection ambiguities in dense pseudo-label methods for detection
• Improving detection performance under category imbalance and dense distributions

Innovation

• LLM-assisted semantic guidance for pseudo-label generation
• Class-Aware Dense Pseudo-Label Assignment for unlabeled data
• Adaptive Hard-Negative Reweighting Module stabilizes learning