RichardScottOZ/mineral-exploration-machine-learning: a curated link list, not a library
List of resources for mineral exploration and machine learning, generally with useful code and examples.
At a glance
- What is it?
- This repository is a README of pointers to prospectivity, remote sensing and geoscience ML projects, organized by domain and author preference. It solves discovery, not modelling, and its value depends on you reading the links rather than expecting runnable code.
- Who is it for?
- Adopt this list if you are a geoscientist or data scientist starting a prospectivity or remote sensing project and need a starting map of where the code lives; the README's own framing is that these are resources the author found useful in practice. Do not adopt it if you need a runnable pipeline, a pinned dependency set, or a supported API, because the repository supplies links and a table of contents rather than installable code.
- Can I use it commercially?
- Not without permission. GitHub finds no licence file in the repository, and without a licence all rights are reserved by default: you may read the code but not reuse it. Check the README, or ask the authors, before using it.
- Is it still maintained?
- Yes. The repository last received commits 39 days ago.
- What is it written in?
- GitHub does not report a main language for this repository.
Answers come from the project's GitHub data, last synced on September 15, 2026, and from our analysis. They are not legal advice.
DEEP OPEN-SOURCE ANALYSIS
What this repository actually is, and the problem it addresses
The README opens by stating the page lists resources for mineral exploration and machine learning, generally with useful code and examples, and adds that these are resources the author has found useful or interesting in practice. That sentence is the whole contract. The problem it addresses is fragmentation: geoscience ML work spans prospectivity mapping, geophysics, geochemistry, spectral unmixing, stratigraphy and natural language processing over reports, and the tooling for each sits in different organizations and different languages. A single index that groups those links by task is a genuine time saver for someone entering the field. It is not a framework, and the repository does not present itself as one. If you clone it expecting an importable package, you will find a README and little else. The audience is narrow and identifiable: people who already know what prospectivity mapping means and need to find the code that does it.
The table of contents is the architecture
There is no module layout to describe, because the repository is documentation. The structure is the README's table of contents, which splits into Prospectivity, Geology, Natural Language Processing, Remote Sensing, Data Quality, Community, Cloud providers, Domains, Overview, Web Services, Data Portals, Tools, Ontologies, Books, Datasets, Papers, Other and General Interest. Below that, a Frameworks section collects the entries that are actual software: UNCOVER-ML and its sub-components, the EIS Toolkit, PySpatialML, DARPA CriticalMAAS and its task repositories, scikit-map, TorchGeo, terratorch, TorchSpatial, torch-harmonics, geodl, Geo Deep Learning, AIDE, ExPLoRA and others. A separate R subsection holds CAST and a geodl R implementation. The grouping is by intent rather than by language or maturity, so a Python library and a paper sit in the same list. That is a reasonable choice for browsing and a poor one for dependency planning, since nothing in the README tells you which entries are maintained.
Forks as a curation signal, and what it costs you
The README states plainly that links currently pointing to a fork are forks where the author changed something to use and put it in the list for reference. That is unusually honest, and it is also the main structural risk of the list. A fork can drift from upstream, can carry local patches that never land, and can stop tracking upstream releases without any visible notice. Several entries follow this pattern, including the UNCOVER-ML Framework, PySpatialML, geo-deep-learning, truly spatial random forest and the Perceiver entries. When you follow one of these links, you are looking at the author's working copy, not the canonical project. For a reference list that is acceptable. For a production dependency it means you should locate the upstream repository and compare before you commit to anything.
Getting it running: there is nothing to install
The material provides no installation command, no package name, no configuration keys and no entry point. There is no setup.py, no pyproject.toml, no environment.yml described, and no release has been retrieved for this repository. The only interaction the README documents is contributing: it invites suggestions through a discussion, issue or pull request. So the honest answer to how you get it running is that you do not run it. You read it, follow a link, and then install that linked project according to its own instructions. The one operational detail worth noting is that at least one link in the CriticalMAAS section is a local file path on the author's own machine, which will not resolve for anyone else. That is a small but telling sign of a personal working document that was published rather than a maintained index.
Where the list stops being useful
The README gives no dates, no version numbers, no maintenance status and no licence information for the linked projects. It also gives no indication of which entries the author has actually used versus which merely looked interesting, despite the framing sentence suggesting a mix of both. The consequence is that dead links, archived repositories and abandoned experiments are indistinguishable from active ones until you click through. There is also no coverage statement: the topics list includes nickel, copper, lithology and stratigraphy, but the README does not claim completeness for any commodity or method, and the presence of a topic tag should not be read as a curated section. If your task is, say, hyperspectral unmixing for a specific deposit type, you may find one or two relevant links and no guidance on which to prefer.
How it differs from a framework like TorchGeo
TorchGeo appears in this list as an entry, and the contrast is instructive. TorchGeo is a PyTorch library for remote sensing style models: you install it, import it, and its datasets and samplers are code you call. This repository is the opposite kind of artifact. It contains no model, no dataset loader and no training loop; it points at projects that do. The difference in approach matters because the two fail differently. A framework fails through version conflicts, API changes and dependency drift, all of which you can pin and test. A link list fails silently through rot, and you only discover it when a URL 404s or a fork has fallen years behind. Neither is better in the abstract. They answer different questions: one asks how do I train this, the other asks what exists that might help me train this.
Maintenance cost and licensing
The repository was last pushed in August 2026 according to the supplied metadata, and no releases have been retrieved. That pattern is consistent with a continuously edited README rather than a versioned project: changes arrive as commits to a single file, and there is no changelog to read. Your maintenance cost as a user is therefore not upgrade work but link triage, done by hand, whenever you revisit the list. On licensing, the repository's own licence is not stated in the material provided, so you cannot assume terms for the README text itself. More importantly, the list aggregates projects under many different licences, and nothing here records which. If you plan to reuse code from a linked repository, check that repository's licence directly. This is a description of what the material does and does not say, not legal advice.
Editorial conclusion
Adopt this list if you are a geoscientist or data scientist starting a prospectivity or remote sensing project and need a starting map of where the code lives; the README's own framing is that these are resources the author found useful in practice. Do not adopt it if you need a runnable pipeline, a pinned dependency set, or a supported API, because the repository supplies links and a table of contents rather than installable code. Before relying on any entry, check the linked repository's own licence, last commit and maintenance status, since this list does not carry that information for you, and note that several links point to forks the author maintains rather than to upstream projects.
Community notes