NN-SVG is a tool for generating publication-ready neural network architecture diagrams as scalable vector graphics.
The tool addresses the time-consuming process of manually drawing neural network schematics by enabling parametric generation instead. It supports three diagram styles: classic fully-connected networks rendered with D3, convolutional networks in the style of LeNet using D3, and deep networks following the AlexNet paper style rendered with Three.js. Users can customize diagrams through numerous parameters controlling size, color, and layout before exporting to SVG format suitable for academic papers and web pages.
Researchers and educators working with neural networks should consider this tool if they need to produce architecture diagrams regularly. It suits projects where visual consistency and publication quality matter, and where manual diagram construction would consume significant time. The tool's parametric approach means changes to network structure can be reflected in diagrams without redrawing from scratch. The README mentions related projects in visualization and network architecture documentation but does not make direct comparisons.
The project shows sparse development activity with infrequent commits and minimal recent updates to the codebase. Issue responses and pull request engagement appear limited, suggesting the maintainer addresses problems on an ad-hoc basis rather than maintaining an active development cadence. The tool remains functional for its core purpose but should not be expected to receive regular enhancements or rapid bug fixes.
