scikit-image is a Python library for image processing that provides a collection of algorithms for analyzing and manipulating images.

A JAX and Flax-based library for computer vision research that provides scalable libraries and project templates for developing attention-based models across images, video, and audio.

A Japanese-specialized Document AI engine that provides high-accuracy OCR, layout analysis, and table structure recognition for images of documents.

InfiniSplat is a system for large-baseline monocular view synthesis that reconstructs 3D Gaussian Splatting scenes from a single RGB image or an RGB-depth pair.

A 3D mesh decomposition tool that breaks complex shapes into approximate convex hulls to enable efficient collision detection and physics simulations.

SplatTransform is an open‑source library and CLI for converting, editing, and analysing 3‑D Gaussian splat data. It supports many input/output formats (PLY, SOG, SPZ, GLB, HTML, voxel, WebP, etc.), offers geometric transforms, filtering, decimation, LOD streaming, collision‑mesh generation, and GPU‑accelerated operations. Install via npm (or use the Docker backend) and run commands like `splat-transform input.ply output.sog` or chain actions (`-s`, `-t`, `--filter-box`, `--decimate`). Ideal for researchers, engine developers, and content creators working with neural‑graphics splat representations.

An open-source DIY golf launch monitor that uses Raspberry Pi and machine learning to track ball launch speed, angles, and spin.

Rust bindings for the OpenCV computer vision library, enabling the use of OpenCV's image processing and CV tools in Rust.

LoMa is a fast, accurate family of local feature matchers for computer vision — it finds corresponding points between image pairs, serving as a drop-in replacement for LightGlue in 3D reconstruction and visual localization pipelines, with multiple model sizes balancing speed and accuracy.

An end-to-end traffic analysis suite that creates 3D digital twins from CCTV footage and satellite maps using computer vision and a custom calibration pipeline.

Sentinel is an open‑source LAN‑only system that turns Android phones into real‑time camera nodes and a Python/Flask PC server into a live preview, segmented recorder, and AI‑triggered monitoring dashboard. It uses motion detection to call external vision models, outputs structured JSON events, and keeps all data local for privacy‑focused monitoring or data‑collection research.

SpatialClaw is a training‑free, code‑as‑action framework for spatial‑reasoning agents. A VLM writes Python cells that run in a persistent Jupyter kernel pre‑loaded with perception tools (SAM‑3, Depth‑Anything‑3) and scientific libraries. The system loops through planning, code generation, safety‑checked execution, and observation until the agent returns an answer. Evaluated on 20 benchmarks, it reaches 59.9 % average accuracy, beating prior agents by 11.2 points, using the same prompt and tools across six VLM back‑ends. The repo provides full runtime, benchmark loaders, installation scripts, and SLURM support.

franken_ocr is a pure‑Rust, CPU‑only OCR engine that runs Baidu’s Unlimited‑OCR and four related vision‑language models. It ships as a single portable binary (`focr`), works completely offline after a one‑time model download, and provides Markdown, JSON, chart‑data, and MusicXML outputs. The project uses hand‑written SIMD kernels, forbids unsafe Rust (except audited SIMD islands), and includes extensive self‑test and release‑certification scripts.

A high-performance 3D Gaussian Splatting and Mesh renderer that uses Chunked Streaming LOD to handle massive 3D datasets.

A repository for experimental and community-contributed extra modules for OpenCV, serving as a testing ground for features before they move to the official core distribution.

A large-scale open-source library for 2D/3D image and point cloud processing, used widely in robotics and 3D perception.

A Rust library and CLI tool for extracting text from images using neural network models to minimize the need for image preprocessing.

DeepLabCut is a markerless pose estimation toolbox that allows researchers to track animals and objects in video without physical markers using deep learning.

A JavaScript library for client-side indecent content detection using TensorFlow.js to classify images into safe and NSFW categories.

AI‑enhanced AntV skill library that lets LLMs generate accurate, ready‑to‑run charts (G2, G6, X6, infographics, narrative text, etc.) via a searchable skill catalogue, CLI, HTTP service, and npm package.

A fast and robust point cloud registration library that aligns 3D point clouds to determine relative rotation and translation.

A state-of-the-art 3D promptable segmentation framework for medical images that allows users to refine volumetric masks using points, scribbles, and lasso prompts.

Track4World is a framework for dense 3D tracking of every pixel in monocular videos, providing world-centric 3D scene flow estimation in a feedforward manner.

A high-performance shot boundary detection tool that uses a Shot-Query Transformer to identify cuts and transitions across diverse video sources.

Face Anything is a feed‑forward neural model that reconstructs a temporally‑consistent 4‑D (3‑D + time) face from any image sequence or single photo, outputting point‑cloud videos, depth/normal maps, canonical coordinate visualisations, per‑frame PLY meshes and camera data. It runs on a CUDA GPU, installs via a one‑line script, and offers configurable processing modes and background removal.

A semantic video search tool that lets you find specific events in footage using natural language or images and automatically trims the matching clips.

A Vulkan RTX path tracer and glTF scene editor that serves as a high-fidelity PBR material reference for glTF 2.0 and 2.1 scenes.

A comprehensive resource hub and benchmark for Infrared and Visible Image Fusion (IVIF), providing curated datasets, a categorized catalog of fusion methods, and evaluation tools.

A cross-domain pre-trained Point Transformer V3 encoder for 3D point clouds, providing a universal representation for various downstream 3D perception tasks.

A cross-platform OpenGL-based 3D game engine featuring extensive procedural generation for universes, cities, and terrain, and high-performance rendering of large models.

A neural network framework for instance segmentation of large, complex 3D shapes, particularly in brain tissue electron microscopy datasets.

A collection of Single Image Super-Resolution models optimized for upscaling and cleaning anime content, offering both real-time and high-quality non-real-time options.

GoCV provides Go language bindings for OpenCV 4, enabling Go developers to integrate advanced computer vision and hardware-accelerated image processing into their applications.

Microsoft’s Global Building Footprints repo releases a >1 billion‑building, open‑licensed dataset of polygon footprints (and many height estimates) generated by deep‑learning segmentation models. The data are provided as gzipped line‑delimited GeoJSON files, indexed by Bing quad‑keys, with confidence scores and documentation for large‑scale GIS or AI‑driven urban analysis.

An encoder-only image segmentation model that repurposes a plain Vision Transformer (ViT) to eliminate the need for complex decoders, offering high speed and accuracy.

WebPlotDigitizer is a computer vision assisted tool that extracts numerical data from images of data visualizations for researchers and scientists.

ReSplat is a feed-forward recurrent model for 3D Gaussian splatting that iteratively refines scene representations using rendering error as feedback.

A pluggable multi-object tracking framework that provides a unified interface for integrating various detectors and trackers for both axis-aligned and oriented bounding boxes.

A computer vision-based Python automation framework for Windows apps, games, and Android devices that supports OCR and template matching.

An AI-powered tool that solves hCaptcha challenges using a mix of computer vision models and multimodal large language models.

A differentiable optical lens simulator for computational imaging and optical design that allows lenses to be optimized as learnable parameters within a PyTorch training loop.

A procedural generation engine that creates infinite photorealistic 3D worlds, indoor scenes, and articulated assets for computer vision and robotics simulation.

D‑FINE is a real‑time object‑detection framework that redefines DETR’s box regression as a fine‑grained distribution refinement and adds a self‑distillation module. It offers several model sizes (‑N to ‑X) that achieve state‑of‑the‑art COCO/AP scores (42.8 %–55.8 %) while running at 70‑470 FPS on a T4 GPU, with no extra inference cost. The repo provides pretrained checkpoints, config files, and full training/evaluation scripts for COCO, Objects365, and custom COCO‑style datasets.

A ComfyUI custom node package that uses GVHMR to extract 3D human motion and SMPL skeletal parameters from video.

A foundation model for zero-shot monocular metric depth estimation that produces high-resolution, sharp depth maps with absolute scale in less than a second.

A target-less, automatic toolbox for LiDAR-camera extrinsic calibration that uses environment structure and texture for precise sensor alignment.

A dependency-free C++ port of Depth Anything 3 and V2 for fast, monocular metric depth and camera pose inference on CPU and GPU.

An Optical Music Recognition (OMR) software that transforms photos or PDFs of sheet music into machine-readable MusicXML files.