ELDOR: A Dataset and Benchmark for Illegal Gold Mining in the Amazon Rainforest

TL;DR

ELDOR introduces a comprehensive UAV dataset with pixel-level annotations for illegal gold mining in the Amazon, enabling new benchmark tasks and revealing current model limitations in detecting small-scale and subtle environmental disturbances.

Contribution

The paper presents ELDOR, a large-scale UAV dataset with detailed annotations, and establishes multiple benchmark tasks for monitoring illegal gold mining, advancing remote sensing and environmental monitoring research.

Findings

Current models struggle with small-scale mining structures.

Fine-grained recovery classes are challenging for existing methods.

Multimodal and context-aware models are needed for improvement.

Abstract

Illegal gold mining in the Amazon rainforest causes deforestation, water contamination, and long-term ecosystem disruption, yet remains difficult to monitor at fine spatial scales. Satellite imagery supports large-scale observation, but often misses small mining-related structures and subtle land-cover transitions, especially under frequent cloud cover. We introduce ELDOR, a large-scale UAV benchmark for monitoring environmental and landscape disturbance from illegal gold mining in the rainforest. ELDOR contains manually annotated orthomosaic imagery covering over 2,500 hectares, with pixel-level semantic labels for both mining-related activities and surrounding ecological structures. With this unified annotation source, we establish four benchmark tasks: semantic segmentation, segmentation-derived recognition, direct multi-label classification, and class-presence recognition with vision-language models. Across these tasks, we compare generic and remote-sensing-specific segmentation models, vision foundation model-related segmentation methods, direct multi-label classification methods, and vision-language models under a controlled closed-set protocol. Results show that current methods still struggle with rare small-scale mining structures and fine-grained recovery classes, suggesting the need for context-aware and multimodal modeling. To support domain analysis and practical use, we further build an interactive explorer for domain experts that provides a unified interface for data exploration and model inference.