Run OpenCLIP compatible embedding models via ONNX Runtime
Object detection using ORT and the yoloe-26-seg model. This model can detect multiple objects per image, each having a tag, pixel-level mask, and a boundingbox. It's pretrained, it has a vocabulary of 4000+ objects.
Pure-Rust AV1 codec — orphan-rebuild scaffold pending clean-room re-implementation.
Pure-Rust H.266 / VVC (ITU-T H.266 | ISO/IEC 23090-3) bitstream parser + decoder scaffold for oxideav
Pure-Rust Fuzzy Inference System (Mamdani + TSK) with PSO optimizer — zero external dependencies
Model implementations for TrustformeRS
Multi-view diffusion model inference for GAF
Zero-dependency NNUE training & inference library in pure Rust
Unified text-to-speech for voice pipelines with multiple backend support
High-performance, deterministic image fingerprinting library
Weight-based backtesting engine for quantitative trading
Local text-to-image, style transfer, LoRA, upscale and color-key CLI built on candle