Projects

### Overview: This project aims to benchmark the scalability, speed, and accuracy of commonly used machine learning algorithms for binary classification. The benchmark focuses on a limited dataset of sizes 10K, 100K, 1M, and 10M, with input features ranging from 1K to 10K. The algorithms studied include linear models (logistic regression, linear SVM), random forests, boosting, and deep neural networks. The primary focus is on evaluating the performance of these algorithms on large-scale datasets, ensuring that the benchmarks can be completed efficiently on commodity hardware. ### Key Features: - **Data Generation:** Training datasets are generated from the airline dataset, using years 2005 and 2006, with a test set of size 100K generated from the same dataset in 2007. - **Scalability and Speed:** Benchmarks are designed to evaluate the scalability and speed of algorithms, with a primary focus on RAM usage. The algorithms are tested on a single machine with decent RAM and multiple cores. - **Accuracy Measurement:** Accuracy is measured using the Area Under the Curve (AUC) metric. - **Commodity Hardware:** The benchmark is designed to run on commodity hardware, ensuring that the results are applicable to real-world scenarios. ### Tech Stack: - **Programming Languages:** R, Python, and various open-source implementations. - **Algorithms:** Linear models (logistic regression, linear SVM), random forests, boosting, and deep neural networks. ### Why It Matters: This benchmark is crucial

### Overview: Guardian is an enterprise-grade AI-powered penetration testing automation framework designed for authorized security testing and educational purposes. It leverages multiple AI providers and security tools to deliver intelligent, adaptive security assessments with comprehensive evidence capture. ### Key Features: - **Multi-Provider AI Intelligence**: Guardian supports 7 AI providers, including OpenAI, Anthropic, Google, OpenRouter, Requesty, Ollama, and OpenAI-compatible models. - **Plugin Provider Contract**: Third-party providers ship via `[project.entry-points."guardian.providers"]` with no need for forks. - **Multi-Agent Architecture**: Specialized agents like Planner, Tool Selector, Analyst, and Judge, along with debate triage roles. - **Multi-Agent Debate Triage**: Red, Blue, and Judge roles debate on ambiguous findings, improving accuracy by 5%. - **Vision-LLM Visual Triage**: Captures headless screenshots and leverages GPT-4o, Claude 3.5+, or Gemini 1.5+ for analyst enrichment. - **RAG Knowledge Base**: SQLite + FTS5 for grounded retrieval over CVE/CWE/Mitre ATT&CK feeds, reducing hallucinated CVE references. - **Judge Model Routing**: `think_deeply` swap-and-restore for big model thinking, small model judging, reducing cost by 10x. - **Learned Tool Selection**: Offline ranker trained on session telemetry, abstains when low-confidence and falls back to LLM selector. - **Adaptive Testing**: AI adjusts tactics based on discovered vulnerabilities and prior tool yields.

Overview: 本项目提供了一个Python库的快速入门教程，专注于Transformer模型及其在自然语言处理中的应用。 Key features: - 示例代码仓库 - 数据集存储在"data"目录中 - 代码分为四个部分：背景知识、初识Transformer、Transformers实战、大语言模型时代 - 包含多个任务示例代码，如句子对分类、序列标注、翻译、文本摘要和情感分析 Tech stack: Python Why it matters: 本教程为自然语言处理领域的开发者提供了一个易于上手的资源，帮助他们快速掌握Transformer模型及其在实际应用中的使用方法。 Repository: jsksxs360/How-to-use-Transformers Primary language: Python GitHub stars: 1896 GitHub description: Transformers 库快速入门教程 README excerpt: 

Overview: NGBoost is a Python library that implements Natural Gradient Boosting, providing a scalable and modular framework for probabilistic prediction. It is built on top of Scikit-Learn and supports various distributions, scoring rules, and base learners. Key features: - Scalable and modular design - Support for different distributions, scoring rules, and base learners - Didactic introduction to the methodology underlying NGBoost Tech stack: Python, Scikit-Learn Why it matters: NGBoost offers a practical and efficient approach to probabilistic prediction, making it suitable for a wide range of applications in machine learning and data science. Repository: stanfordmlgroup/ngboost GitHub stars: 1896 GitHub description: Natural Gradient Boosting for Probabilistic Prediction

### Overview: This repository is a curated list of radar datasets, detection, tracking, and fusion techniques, focusing on autonomous driving applications. It serves as a valuable resource for researchers and practitioners in the field of radar perception. ### Key Features: - **Datasets**: Comprehensive collection of radar datasets including nuScenes, DENSE, PixSet, Radar Scenes, Pointillism, Zendar SAR, and Cooperative Radars. - **Detection and Tracking**: Techniques for detecting and tracking objects using radar data. - **Fusion**: Methods for combining radar data with other modalities to improve accuracy and reliability. - **Augmentation**: Tools and techniques for enhancing radar data for better performance. ### Tech Stack: - Python, MATLAB, and other relevant software tools for data processing and analysis. ### Why It Matters: This repository is crucial for researchers and practitioners in autonomous driving, offering a wide range of datasets and techniques that can significantly enhance the performance of radar-based perception systems. It serves as a valuable resource for developing and improving radar-based autonomous driving technologies. Repository: ZHOUYI1023/awesome-radar-perception GitHub Stars: 1895

Overview: TensorRT-YOLO是一款专为NVIDIA设备设计的易用灵活、极致高效的YOLO系列推理部署工具。它集成了TensorRT插件以增强后处理效果，并使用CUDA核函数和CUDA图加速推理过程。项目提供了C++和Python推理的支持，旨在提供开箱即用的部署体验，涵盖目标检测、实例分割、图像分类、姿态识别、旋转目标检测和视频分析等任务场景。 Key features: - **CUDA 加速**：通过CUDA核函数优化前处理流程，并采用CUDA图技术加速推理过程。 - **TensorRT 集成**：深度集成TensorRT插件，显著加速后处理，提升整体推理效率。 - **多 Context 推理**：支持多Context并行推理，最大化硬件资源利用率。 - **显存管理优化**：适配多架构显存优化策略（如Jetson的Zero Copy模式），提升显存效率。 - **开箱即用**：提供全面的C++和Python推理支持，满足不同开发者需求。 - **CLI工具**：提供命令行界面简洁直观，并支持自动识别模型结构，无需复杂配置。 - **Docker支持**：提供Docker一键部署方案，简化环境配置与部署流程。 Tech stack: C++ / Python Why it matters: TensorRT-YOLO旨在提供一种简单、高效且灵活的解决方案，帮助开发者快速部署YOLO模型，满足各种应用场景的需求。通过CUDA和TensorRT的集成，显著提升了推理性能，

Overview: Petastorm is a Python library that enables training and evaluation of deep learning models directly from Apache Parquet datasets. It supports popular ML frameworks such as TensorFlow, PyTorch, and PySpark, and can be used from pure Python code. Key features: - Supports TensorFlow, PyTorch, and PySpark - Enables single machine or distributed training and evaluation - Supports Apache Parquet format for dataset storage - Extensible data codecs for custom compression - PySpark natively supports Parquet format for easy integration with Spark Tech stack: Python Why it matters: Petastorm simplifies the process of training and evaluating deep learning models on large datasets, making it easier to scale and deploy models in production environments.

Overview: Navop is a comprehensive workspace designed for databases, SSH, SFTP, terminals, remote desktop, monitoring, and AI agents. Built with GPUI and Rust, it offers GPU-accelerated rendering without the need for a WebView. Key features: - Native, all-in-one workspace - Databases (MySQL, PostgreSQL, SQLite, DuckDB, ClickHouse, SQL Server, Oracle, TDengine, Dameng DM, KingbaseES, GBase 8a, GBase 8s, OceanBase, Apache IoTDB) - SSH/SFTP/FTP/FTPS/Port Forwarding/RDP/VNC - Monitoring and AI agents Tech stack: Rust Why it matters: Navop simplifies access to multiple computing environments, providing a unified interface for databases, remote connections, and monitoring tools, making it ideal for developers and IT professionals looking to streamline their workflow.

Overview: gplearn is a Python library that implements Genetic Programming, a technique inspired by the scikit-learn API, for solving symbolic regression problems. Key features: - Symbolic regression - Symbolic classification - Feature engineering transformation Tech stack: Python Why it matters: gplearn provides a powerful and flexible tool for symbolic regression, leveraging the scikit-learn API for ease of use and compatibility with existing machine learning workflows.

Overview: TuyaOpen is a cutting-edge AI+IoT framework designed for developing next-gen hardware products, including Tuya T-Series MCUs, Pi, and ESP32s. It integrates with leading LLMs and AI platforms, enabling advanced multimodal AI capabilities. Key features: - Supports cross-platform C/C++ SDK for gear (Tuya T-Series, Pi, ESP32s) - Integrates with Tuya Cloud for low-latency multimodal AI - Pairs with Tuya Cloud’s AI workflows for drag-and-drop integration - Streamlines open AI-IoT ecosystem building - Develops smart devices with voice, vision, and sensor-based features - Seamlessly connects to Tuya Cloud for remote control, monitoring, and OTA updates - Compatible with Google Home and Amazon Alexa - Custom hardware with the "Powered by Tuya" label - Supports Bluetooth, Wi-Fi, and ET Tech stack: C/C++ Why it matters: TuyaOpen simplifies the development of smart devices by integrating advanced AI capabilities, making it ideal for creating intelligent hardware solutions across various applications. Repository: tuya/TuyaOpen GitHub stars: 1892

Overview: This is an unofficial C#/.NET SDK for accessing the OpenAI GPT-3 API, including GPT-3.5/4, GPT-3.5/4-Turbo, and DALL-E 2/3. It's based on .NET Standard 2.0 and supports various platforms and environments. Key features: - Easy to use wrapper for OpenAI's API - Supports multiple versions of GPT-3 and DALL-E - Provides comprehensive documentation and examples - Compatible with various .NET frameworks and platforms Tech stack: C# Why it matters: This SDK is useful for developers who need to integrate OpenAI's advanced AI capabilities into their projects without having to rely on the official library. It offers a simple and reliable way to access the latest versions of GPT-3 and DALL-E, making it easier to implement and test. Repository: OkGoDoIt/OpenAI-API-dotnet GitHub stars: 1892 GitHub description: An unofficial C#/.NET SDK for accessing the OpenAI GPT-3 API, including GPT-3.5/4, GPT-3.5/4-Turbo, and DALL-E 2/3. It's based on .NET Standard 2.0 and supports various platforms and environments.

Overview: This repository contains TensorFlow binaries optimized for AVX, FMA, and SSE architectures, making it suitable for performance-critical applications. Key features: - Optimized TensorFlow binaries for AVX, FMA, and SSE - Supports advanced SIMD instructions for improved performance Tech stack: Shell Why it matters: This repository is essential for developers looking to enhance the performance of TensorFlow applications by leveraging advanced SIMD instructions, thereby reducing latency and improving overall system efficiency.

Overview: This repository is a PyTorch implementation of various methods for continual learning, including XDG, EWC, SI, LwF, FROMP, DGR, BI-R, ER, A-GEM, iCaRL, and Generative Classifier, across three different scenarios. Key features: - Supports experiments in the academic continual learning setting. - Includes scripts for running both task-based and task-free continual learning scenarios. - Provides a NeurIPS tutorial and ICLR blog post for further details and instructions. Tech stack: PyTorch and Torchvision. Why it matters: This repository offers a comprehensive solution for conducting continual learning experiments, making it valuable for researchers and practitioners in the field of machine learning.

### Overview: imgcook is an image-to-code generator that intelligently converts design documents into front-end codes, enhancing collaboration and development efficiency between designers, developers, and testers. ### Key Features: - **One-click Restoration of Design Drafts**: Automatically restore design drafts to ensure consistency and accuracy. - **Visual Editing**: Edit design views, logic, and other components directly within the editor. - **Code Export**: Export generated code to local or project files for further development. - **Engineering Efficiency Tools**: Includes imgcook-cli, VS Code plugins, and support for custom DSL and plugins. - **AI Capabilities**: Supports image to code conversion and text semantic understanding. - **Team and Project Management**: Manage modules by team and project dimensions. ### Tech Stack: Node.js, TypeScript, React, and other related technologies. ### Why It Matters: - **Improved Development Efficiency**: Enhances collaboration and development efficiency by automating the restoration and code generation process. - **Enhanced Code Quality**: Ensures that the generated codes are consistent with the design drafts, reducing the need for manual corrections. - **Increased Development Speed**: Facilitates rapid development of applications, especially in scenarios where quick delivery is crucial. ### Repository: imgcook/imgcook ### GitHub Stars: 1890 ### GitHub Description: Generate pages from any sketch or images.

### Overview: 车牌识别技术是现代交通监控和智能分析的重要组成部分。本项目利用YOLOv5模型实现中文车牌的检测与识别，支持多种车牌类型，包括12种中文车牌和双层车牌。项目提供详细的车牌检测和识别训练数据集，适用于各种场景下的车牌检测和识别需求。 ### Key features: - 支持12种中文车牌类型 - 支持双层车牌 - 车牌检测与识别功能 - 提供车牌检测训练和车牌识别训练的链接 - 支持多种车牌类型，如单行蓝牌、单行黄牌、新能源车牌、警用车牌等 - 提供图片测试和视频测试demo，方便用户快速验证功能 ### Tech stack: Python ### Why it matters: 车牌识别技术在智能交通系统中具有广泛的应用前景。通过识别车牌，可以实现车辆的自动识别、监控和管理，提高交通管理的效率和安全性。本项目提供的车牌检测与识别功能，能够满足各种场景下的车牌识别需求，具有很高的实用价值。 ### Repository: Repository: we0091234/Chinese_license_plate_detection_recognition GitHub stars: 1890 GitHub description: yolov5 车牌检测 车牌识别 中文车牌识别 检测 支持12种中文车牌 支持双层车牌 ### Environment requirements: - Python >=3.6 - PyTorch >=1.7 ###

Overview: This project is an implementation of the Neural Collaborative Filtering (NeuMF) model, designed for implicit feedback and ranking tasks. It targets three collaborative filtering models: Generalized Matrix Factorization (GMF), Multi-Layer Perceptron (MLP), and Neural Matrix Factorization (NeuMF). Key features: - Neural Collaborative Filtering model - Optimized using log loss with negative sampling - Supports implicit feedback and ranking tasks - Available in Python and Keras with Theano backend - Supports pre-training of NeuMF model Tech stack: Python, Keras, Theano Why it matters: The implementation provides a practical tool for researchers and developers interested in improving collaborative filtering models, particularly in scenarios where implicit feedback is prevalent. Repository: hexiangnan/neural_collaborative_filtering Primary language: Python GitHub stars: 1888 GitHub description: Neural Collaborative Filtering

**Overview:** This repository contains a collection of Core ML models and sample apps, providing a comprehensive library for iOS and macOS developers. The models are converted from PyTorch and are ready to use in Swift projects. The models are categorized into various categories such as image classification, object detection, multi-object tracking, segmentation, video matting, super resolution, low light enhancement, image restoration, image generation, and image2image. **Key Features:** - CoreML models for image classification, object detection, multi-object tracking, segmentation, video matting, super resolution, low light enhancement, image restoration, image generation, and image2image. - Sample apps demonstrating how to integrate these models into Swift projects. - Conversion scripts for models from PyTorch to Core ML format. - Sample projects for each model type, including `xLights` for an open-source light show sequencer. **Tech Stack:** Swift **Why it Matters:** This repository serves as a valuable resource for developers looking to integrate state-of-the-art AI models into their iOS and macOS projects. The models are easy to use and integrate, making it a go-to solution for developers seeking to enhance their applications with AI capabilities.

Overview: ReCamMaster is a camera-controlled generative rendering framework that re-captures in-the-wild videos using novel camera trajectories. It achieves this by employing a simple yet effective video conditioning scheme. Key features: - Re-captures videos using novel camera trajectories - Released MultiCamVideo dataset rendered with Unreal Engine 5 - Open-source repository with detailed documentation and project page Tech stack: Python Why it matters: ReCamMaster provides a practical solution for re-capturing videos in real-world scenarios, offering a valuable tool for researchers and practitioners in the field of video synthesis and generation.

### Overview: nndeploy 是一款简单易用和高性能的 AI 部署框架，旨在解决 AI 算法在端侧部署的问题。它支持桌面端（Windows、macOS）、移动端（Android、iOS）、边缘计算设备（NVIDIA Jetson、Ascend310B、RK 等）以及单机服务器（RTX 系列、T4、Ascend310P 等）的部署。通过可视化工作流和多端推理，实现 AI 算法在上述平台和硬件上的高效、高性能落地。 ### Key features: - **可视化工作流**：拖拽节点即可部署 AI 算法，参数实时可调，效果一目了然。 - **自定义节点**：支持 Python/C++自定义节点，无论是用 Python 实现预处理，还是用 C++/CUDA 编写高性能节点，均可无缝集成到与可视化工作流。 - **一键部署**：工作流支持导出为 JSON，可通过 C++/Python API 调用，适用于 Linux、Windows、macOS、Android 等平台。 ### Tech stack: C++ ### Why it matters: nndeploy 适合作为一款可视化工作流工具，特别适用于处理 10B 以上的大模型（如大语言模型和 AIGC 生成模型），提供高效、高性能的 AI 算法部署解决方案。

Overview: CTransformers is a Python library that provides bindings for the Transformer models implemented in C/C++ using the GGML library. This library simplifies the integration of these models into Python applications, making them accessible and efficient for developers. Key features: - 🤗 Transformers integration: Allows seamless use of Transformers models with CTransformers. - GGML model support: Supports models implemented in C/C++ using the GGML library. - GPU acceleration: Enables running some model layers on the GPU for improved performance. Tech stack: C/C++ and Python. Why it matters: CTransformers simplifies the integration of Transformer models into Python applications, making them more accessible and efficient for developers, especially those working with large language models. This library is particularly useful for applications that require high performance and flexibility in handling large language models.