RESIDUALS
Feature detection in terrain data through systematic decomposition analysis
Visit residuals.briansheppard.com ↗What is RESIDUALS?
RESIDUALS is a framework for analyzing Digital Elevation Models by systematically combining 25 decomposition methods with 19 upsampling techniques to reveal terrain features invisible in raw data. It tests approximately 40,000 algorithm combinations across parameter ranges, generating 39,731 residual outputs totaling 4.28 TB of data. The tool is designed for archaeological prospection and general terrain analysis.
The project provides multiple ways to explore results: an interactive 3D UMAP atlas embedding algorithm signatures, a visual grid comparing decomposition methods side-by-side, parameter-sweep videos showing how individual settings affect output, and a Blender visualization of the embedding space. The maker built it with Cursor and Opus 4.5, using NumPy, SciPy, scikit-image, PyWavelets, and OpenCV. Code and datasets are available on GitHub and Hugging Face.
Momentum
Daily activity over the last 30 days: upvotes, unique visits, new GitHub stars and Hacker News points.
How do we know it's vibe coded?
vibetrend discovered RESIDUALS automatically. Evidence: HN post: "Built in Cursor with Opus 4.5, NumPy, SciPy, scikit-image". View source.
Similar vibe-coded apps
Open-source Wyckoff trading agent for volume-price analysis across A-shares
Open-source Java library for fluid behavior and process simulation
Open skill registry and toolkit for AI agents to build reliable data products
Open specification for decentralized data pods under user control
AI-powered GSoC organization finder with historical data and analytics
Open-source control plane for orchestrating multiple AI coding agents
Open-source AI job search agent that runs locally in your CLI