ModLens is a plug‑in for DeepSeek Harness (and other local AI assistants) that adds vision capability to text‑only LLMs. Users paste an image directly into the chat; the plugin forwards it to a configurable vision engine (Gemini, OpenAI‑compatible, Anthropic, Antigravity CLI, Claude/Kimi CLIs, or local agents) and returns a structured transcription for the model to cite. Installation is a single `npx` command or via `skills.sh`, with zero‑config defaults and automatic key rotation/fail‑over. The project is MIT‑licensed and focuses on lightweight integration rather than heavy wrappers.

An OBS Studio plugin that uses neural networks to remove portrait backgrounds and enhance low-light video in real-time without requiring a physical green screen.

A real-time object detection framework based on the Detection Transformer (DETR) architecture that aims to outperform YOLO models in speed and accuracy.

An automation assistant for the game Endfield that uses screen recognition to automate combat, resource gathering, and daily management tasks.

MoGe is a model for recovering accurate 3D geometry, including metric depth and point maps, from single open-domain images.

An advanced 3D Gaussian Splatting renderer for THREE.js that enables high-performance, cross-device rendering of splat-based 3D scenes.

An image-recognition-based automation framework for black-box testing that allows developers to create software assistants and automation tools using a low-code pipeline protocol.

A multi-channel video analysis platform that combines YOLO detection and LLM verification to provide structured alarms for security and monitoring.

A privacy-preserving human pose estimation system that uses WiFi Channel State Information (CSI) and machine learning to detect poses without cameras.

A CUDA-accelerated library for the rasterization of 3D Gaussians, providing a faster and more memory-efficient way to render radiance fields compared to the official implementation.

A reverse-engineering project that detects and removes Google's invisible SynthID watermarks from Gemini-generated images using spectral analysis and a multi-stage adversarial pipeline.

An automated tool for segmenting major anatomical structures in CT and MR images, providing detailed anatomical masks and body statistics.

A Windows desktop assistant for League of Legends ARAM that provides champion selection advice and uses OCR to recommend builds based on in-game Hextech Augments.

An implementation of the AlphaFold 3 inference pipeline for the accurate 3D structure prediction of biomolecular interactions.

Open3D is a modern library for 3D data processing that provides optimized tools for point cloud and mesh manipulation, visualization, and 3D machine learning.

A deep learning library for benchmarking and deploying anomaly detection algorithms, with a strong focus on visual anomaly detection in images and videos.

An open-source Android document scanner that uses local machine learning for automatic document detection and perspective correction to create private PDFs.

A free and open-source motion capture platform that provides research-grade tracking without requiring expensive, proprietary hardware.

A computer vision framework for pretraining, fine-tuning, and distilling SOTA models like DINOv3 and LTDETRv2 for tasks such as object detection and segmentation.

A comprehensive PyTorch library providing a vast collection of state-of-the-art computer vision backbones and pretrained weights for easy integration into AI projects.

A commercial-grade software for processing drone images into georeferenced maps, point clouds, 3D models, and Gaussian splats.

BiRefNet is a high-resolution dichotomous image segmentation model that provides state-of-the-art background removal and object segmentation for images of varying resolutions.

LLM Compressor is a Python library that quantizes and prunes large language models into the `compressed‑tensors` format, enabling memory‑efficient deployment with vLLM. It supports many low‑precision formats (NVFP4, FP8, INT4, etc.), several PTQ/GPTQ algorithms, DDP and disk‑offloading for huge models, and ships pre‑quantized checkpoints for popular LLMs.

The NVIDIA DLSS SDK provides tools for high-performance image scaling and sharpening for games, supporting integration with DX11, DX12, and Vulkan.

A Python package that integrates AI frameworks with geospatial data analysis, providing tools for satellite imagery processing, model training, and inference.

A cross-platform SDK for RealSense depth cameras that enables depth and color streaming and provides calibration data for computer vision and robotics.

An open-source tool for building high-quality computer vision datasets and models by providing visual data exploration, labeling, and model evaluation.

A family of real-time stereo matching architectures that achieve strong zero-shot generalization by using knowledge distillation, neural architecture search, and structured pruning.

A comprehensive benchmark for evaluating document parsing in real-world scenarios, featuring rich annotations for layout detection, OCR, table, and formula recognition.

SHARP is a fast monocular view synthesis tool that uses a neural network to generate a metric 3D Gaussian representation from a single image in under a second.

A powerful and flexible codebase for point cloud perception research, integrating various 3D backbones and self-supervised pre-training frameworks.

A family of conditional generative models that repurpose pretrained diffusion models like Stable Diffusion for high-resolution monocular depth, surface normal, and intrinsic image analysis.

A plug-and-play Python library providing benchmarked implementations of multiple object tracking algorithms for any detection model.

A node-based visual programming framework for 3D reconstruction and computer vision, enabling users to build complex data processing pipelines using AI and photogrammetry.

An AI-powered NVR system that adds object detection, tracking, and AI-generated event summaries to any RTSP security camera.

A YOLO-style framework for real-time object detection that uses Mixture-of-Experts (MoE) to adaptively allocate computation based on scene complexity for higher precision and lower latency.

An AI-powered image annotation tool that integrates YOLOv8 and Segment Anything to provide automatic labeling for computer vision datasets.

An open-source research framework for universal metric 3D reconstruction that uses a single transformer model to handle multiple 3D tasks from various input combinations.

A video cut detection and analysis tool that automatically identifies scene changes to split videos into individual shots or extract frames.

A family of parametric statistical 3D human models, starting with GNM Head, for high-fidelity representation of human geometry and appearance.

A differentiable computer vision library for PyTorch that integrates classical image processing and geometric vision algorithms into deep learning pipelines.

An open-source Optical Music Recognition (OMR) application that transcribes music score images into symbolic digital formats for playback and editing.

A 650M parameter OCR model for document intelligence that provides high-accuracy text recognition, layout analysis, and table extraction across 91 languages.

A GPU-accelerated MediaPipe plugin for TouchDesigner that enables real-time computer vision tasks like pose and hand tracking without requiring external installations.

An open-source face recognition SDK for Windows and Linux that provides on-premise face detection, landmark detection, and similarity comparison using deep learning.

An automation tool for the game *異環* (The Ring) that uses image and audio recognition to automate repetitive gameplay and mini-games.

A high-performance, cross-platform C++ library for image and video beauty filters using OpenGL/ES.

TESSERA is an open geospatial foundation model that converts cloud-corrupted satellite time series into compact 128-dimensional embeddings for global Earth representation and analysis.