Gold Exploration using Representations from a Multispectral Autoencoder

2026-02-06

AI Summary

Gold exploration is constrained by the high cost and scarcity of in-situ data, necessitating scalable remote sensing approaches. This work proposes an efficient prospecting framework leveraging Sentinel-2 multispectral imagery, wherein high-information-density spatio-spectral representations are extracted using an Isometric autoencoder pretrained on FalconSpace-S2 v1.0 and subsequently fed into a lightweight XGBoost classifier to identify gold-bearing regions. To our knowledge, this is the first application of generative representations from a multispectral autoencoder foundation model to gold mineralization prediction, effectively capturing transferable mineralogical patterns under limited labeled data. Experimental results demonstrate significant improvements over raw spectral baselines, with patch-level and image-level accuracies increasing from 0.51 and 0.55 to 0.68 and 0.73, respectively, across 63 test images.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Computer Vision: Remote Sensing / Geospatial AI

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Satellite imagery is employed for large-scale prospectivity mapping due to the high cost and typically limited availability of on-site mineral exploration data. In this work, we present a proof-of-concept framework that leverages generative representations learned from multispectral Sentinel-2 imagery to identify gold-bearing regions from space. An autoencoder foundation model, called Isometric, which is pretrained on the large-scale FalconSpace-S2 v1.0 dataset, produces information-dense spectral-spatial representations that serve as inputs to a lightweight XGBoost classifier. We compare this representation-based approach with a raw spectral input baseline using a dataset of 63 Sentinel-2 images from known gold and non-gold locations. The proposed method improves patch-level accuracy from 0.51 to 0.68 and image-level accuracy from 0.55 to 0.73, demonstrating that generative embeddings capture transferable mineralogical patterns even with limited labeled data. These results highlight the potential of foundation-model representations to make mineral exploration more efficient, scalable, and globally applicable.

Problem

• gold exploration
• prospectivity mapping
• satellite imagery
• mineral exploration
• multispectral data

Innovation

• multispectral autoencoder
• foundation model
• generative representation
• mineral exploration
• Sentinel-2 imagery