TriangleVision

A high-performance video processing tool that converts standard video into a stylized mesh of triangles in real-time.

View on GitHub
0 / 0

Core Features

📐

Delaunay Triangulation

Advanced geometric algorithms create a high-fidelity mesh that preserves the structural integrity of your video.

Real-time Processing

Optimized for 30FPS+ at 720p resolution using GPU-accelerated rendering and threaded capture.

👁️

Intelligent Sampling

Adaptive point sampling based on edge density, human detection (HOG/Haar), and motion tracking.

📦

Custom .triv Codec

A highly efficient binary format designed specifically for triangle geometry and color data.

🎨

Artistic Styles

Switch between standard triangles, rotoscope mode (ink edges + cell shading), and heatmap visualizations.

🎥

Full Pipeline

Tools for live webcam playback, batch encoding from standard formats, and high-quality MP4 exports.

Engineered for Performance

ModernGL (GPU)
Numba (JIT)
NumPy Vectorization
Throttled Detection
Threaded IO
SciPy Spatial

By leveraging native libraries and GPU acceleration, TriangleVision minimizes CPU overhead while maximizing triangle throughput.