Category
Explore by categories
Curated collection of production-oriented AI projects that implement OCR, RAG, multi-agent systems, and multimodal pipelines. Each entry provides runnable code, setup notes, and engineering patterns to help developers move prototypes toward production.
Real-time DETR detector on a DINOv2 backbone, covering detection, segmentation, and keypoints. Ships in six sizes (Nano to 2XL), beats YOLO on the COCO speed-accuracy curve, and transfers better to non-COCO real-world domains.
Converts document images—scans, photos, born-digital PDFs—into structured text in two stages: first map layout and reading order, then parse each element (text, tables, formulas, figures) in parallel, each guided by its own task prompt.
Real-time 3D Gaussian Splatting renderer for web apps using THREE.js. Integrates splat and mesh rendering with a Rust + Wasm component, supports major splat formats (.PLY, .SPZ, .SOG) and targets broad WebGL2 support for mobile-friendly dynamic scenes.
GPU-accelerated, non-destructive RAW photo editor designed for fast, low-footprint desktop workflows (<20MB). Built with Rust/WGPU/Tauri, it offers real-time 32-bit GPU processing, AI-assisted masking and optional ComfyUI integration for generative edits, plus presets and batch export.
Detects, segments, and tracks every instance of an open-vocabulary concept in images and video from a text phrase or visual exemplar, not just one object per prompt. An 848M-param model reaching ~75-80% of human accuracy across 270K concepts.
A 20B-parameter MMDiT diffusion model that generates and edits images with accurate embedded text, including dense Chinese and English typography. Handles complex multi-line layouts and identity-preserving edits while keeping text legible.
Self-supervised vision foundation model producing dense, patch-level features that transfer to classification, segmentation, depth, and detection with a frozen backbone. Spans ViT-S (21M) to ViT-7B (6.7B params), plus ConvNeXt and satellite variants.
Converts images and PDFs into structured Markdown, HTML, or JSON while preserving layout, handling tables, math, handwriting, charts, and chemistry diagrams across 90+ languages. Runs locally via HuggingFace or against a vLLM server.
Turn plain-English requests into editable draw.io diagrams: the model writes the underlying draw.io XML, which renders live in an embedded canvas. Upload images, PDFs, or text to replicate, refine through chat, and roll back via version history.
Generates real-time, infinite-length portrait video from one reference image on a 12GB GPU. Combines implicit facial signals and 3D keypoints with step-distilled diffusion and autoregressive micro-chunk streaming for low-latency live use.