Projects

### Overview: This repository is a Semantic Segmentation Suite designed to simplify the implementation, training, and testing of various Semantic Segmentation models using TensorFlow. It includes training and testing modes, data augmentation, support for multiple state-of-the-art models, flexibility with any dataset, and comprehensive evaluation metrics. ### Key Features: - **Training and Testing Modes**: Facilitates easy setup for both training and testing of Semantic Segmentation models. - **Data Augmentation**: Enhances model robustness through various data augmentation techniques. - **State-of-the-Art Models**: Includes popular and advanced models like SegNet, ResNet, and others. - **Plug-and-Play**: Allows seamless integration with different models without modification. - **Versatile Data Support**: Compatible with any dataset, ensuring broad applicability. - **Evaluation Metrics**: Provides precision, recall, F1 score, average accuracy, per-class accuracy, and mean IoU for comprehensive performance assessment. ### Tech Stack: TensorFlow ### Why It Matters: This repository is essential for researchers and developers working on Semantic Segmentation projects, offering a comprehensive solution for training, testing, and evaluating various models efficiently. Its flexibility and ease of use make it a valuable tool for anyone looking to improve their Semantic Segmentation capabilities.

Overview: This repository contains 250+ detailed flashcards designed for reviewing topics in machine learning, computer vision, and computer science. The flashcards are useful for self-study and interview preparation. Key features: - Topics covered include computer science, classical ML, modern deep learning, 2D/3D computer vision, NLP, reinforcement learning, generative models, and more. - Updated with topics in RL, NeRFs, Gaussian splatting, generative models, LLMs, and VLMs. - Switched to PowerPoint for better equation editing. Tech stack: PowerPoint. Why it matters: The flashcards provide a comprehensive overview of various machine learning topics, including recent advancements and techniques. They can serve as a valuable reference for both self-study and interview preparation.

Overview: This repository is a comprehensive collection of papers, datasets, code, and other resources for object tracking and detection using deep learning. It serves as a valuable resource for researchers, developers, and practitioners looking to enhance their skills in this field. Key features: - Papers on object tracking and detection using deep learning - Datasets for training and testing models - Code implementations of various tracking and detection algorithms - Documentation and tutorials for beginners Tech stack: Python, TensorFlow, OpenCV, and other relevant deep learning libraries Why it matters: This repository is essential for anyone interested in advancing their knowledge and skills in object tracking and detection using deep learning. It provides a wealth of resources that can be used to develop and improve tracking and detection algorithms, making it a must-have for professionals in the field.

Overview: CodeMachine is an open-source tool that orchestrates AI coding agents into repeatable, long-running workflows. Key features: - Captures workflows from interactive to autonomous execution - Supports multi-agent orchestration and parallel execution - Handles long-running workflows and context management - Enables developers to build complex workflows once and run them reliably Tech stack: TypeScript Why it matters: CodeMachine simplifies the process of building and executing complex workflows, reducing time and effort on repetitive tasks and enabling developers to focus on the core functionality of their projects.

Overview: EasyLM is a comprehensive solution for training, finetuning, evaluating, and serving large language models using JAX/Flax, making it accessible and scalable. Key features: - EasyLM simplifies the process of training, finetuning, evaluating, and serving large language models with JAX/Flax. - Supports pre-training, finetuning, and serving of models like LLaMA, LLaMA 2, and LLaMA 3. - Leverages JAX's pjit functionality to train models on multiple accelerators, including TPU and GPU. - Built with JAX/Flax, providing an easy-to-use and customizable codebase. - Supports multiple models, including OpenLLaMA and Koala chatbots. Tech stack: JAX/Flax Why it matters: EasyLM simplifies the complex process of training large language models, making it accessible and scalable, especially for those looking to train models on multiple accelerators without the complexity of other frameworks.

### Overview: Claude Scientific Writer is a powerful tool that combines AI-driven deep research with well-formatted written outputs, enabling the creation of publication-ready scientific papers, reports, posters, grant proposals, literature reviews, and more. It ensures every claim is supported by real, verifiable sources, making it a valuable resource for researchers and writers alike. ### Key features: - **Real-time Literature Search**: Ensures every claim is backed by real, verifiable sources. - **Comprehensive Research**: Performs comprehensive research before writing, ensuring accuracy. - **AI-Powered Diagram Generation**: Generates diagrams and figures for visual representation. - **Document Conversion**: Converts documents into formats like PDF, DOCX, and PPTX. - **Parallel Search and Extract**: Supports parallel search and extract for faster research. - **Python Package**: Available as a Python package for easy integration. - **CLI**: Command-line interface for quick access. - **Claude Code Plugin**: Integrates seamlessly with Claude Code for enhanced functionality. ### Tech stack: The tool is built using Python, leveraging libraries such as Anthropic, Parallel Search and Extract, and OpenRouter for AI image generation. It supports LaTeX for PDF generation and optional LibreOffice for Office rendering and conversion. ### Why it matters: Scientific Writer is essential for researchers and writers looking to produce high-quality, publication-ready documents. It ensures accuracy and credibility by combining AI-driven research with well-formatted outputs, making it a valuable resource in the academic and scientific community. Repository: K-Dense-AI/claude-scientific-writer

Overview: Mantis is a modular toolkit designed to help AI-driven agents find, reproduce, and patch vulnerabilities in software, ensuring robust security practices in the AI-driven software development era. Key features: - AI-driven vulnerability detection - Autonomous code generation and execution - Vulnerability patching and reporting - Secure software development practices Tech stack: Python Why it matters: Mantis provides a comprehensive solution for developers to implement secure software practices using AI, ensuring that vulnerabilities are identified, patched, and validated by human experts, thereby reducing the risk of security breaches and maintaining the integrity of the software ecosystem.

### Overview: This repository contains popular and reproducible single image denoising works, inspired by the comprehensive summary by flyywh. It includes a collection of state-of-the-art algorithms for image denoising, focusing on filtering and sparse coding methods. The repository provides codes and reproducible results that demonstrate the effectiveness of these techniques. ### Key Features: - **Filtering Algorithms:** - **NLM (Non-Local Means Filter):** Image denoising based on non-local means filter and its method noise thresholding. - **BM3D (Block Matching and 3D Filtering):** Image restoration by sparse 3D transform-domain collaborative filtering. - **PID (Progressive Image Denoising):** Progressive image denoising with sparse and redundant representations over learned dictionaries. - **Sparse Coding Algorithms:** - **KSVD (K-SVD):** Image denoising via sparse and redundant representations over learned dictionaries. - **LSSC (Local Sparse Coding):** Non-local sparse models for image restoration. - **NCSR (Non-Centralized Sparse Representation):** Nonlocally centralized sparse representation for image restoration. - **OCTOBOS (Overcomplete Structured Sparse Coding):** Structured overcomplete sparsifying transform learning with convergence guarantees and applications. - **GSR (Group-Based Sparse Representation):** Group-based sparse representation for image restoration. - **TWSC (Trilateral Weighted Sparse Coding):** A trilateral weighted sparse coding scheme for real-world image denoising. ### Tech

### Overview: This repository provides a clean implementation of YOLOv3 in TensorFlow 2.0, including pre-trained weights, inference examples, transfer learning, and more. It is designed to be fully integrated with `absl-py` and follows best practices. ### Key Features: - TensorFlow 2.0 - Pre-trained Weights - Inference Example - Transfer Learning Example - Eager Mode Training - Graph Mode Training - Functional Model - Input Pipeline - Tensorflow Serving - Vectorized Transformations - GPU Acceleration - Clean Implementation - MIT License ### Tech Stack: TensorFlow 2.0 ### Why it Matters: This implementation is designed for developers and researchers looking to integrate YOLOv3 into their projects using TensorFlow 2.0, ensuring compatibility and efficiency. The clean and best-practice implementation makes it easy to start from scratch or fine-tune existing models.

Overview: This repository offers a comprehensive set of Jupyter Notebook tutorials on applying Deep Learning models such as NLP, Computer Vision, and Time Series forecasting using PyTorch. Key features: - Real-world problem-solving with Deep Learning models - Prototyping to deployment with PyTorch and Python - Tutorials on Face detection with Detectron 2, Time Series anomaly detection with LSTM Autoencoders, Object Detection with YOLO v5, and more Tech stack: Jupyter Notebook Why it matters: These tutorials provide practical, hands-on experience in applying Deep Learning models to real-world problems, making the transition from theory to practice seamless. Repository: curiousily/Getting-Things-Done-with-Pytorch Primary language: Jupyter Notebook GitHub stars: 2513 GitHub description: Jupyter Notebook tutorials on solving real-world problems with Machine Learning & Deep Learning using PyTorch.

### Overview: CodeIsland is a real-time AI coding agent status panel that displays live status, approvals, and replies for 30+ AI coding tools on your MacBook notch. It also has iPhone and Apple Watch companions, making it accessible and convenient. ### Key Features: - **Live Status and Replied Sessions**: Show the current session, latest reply, and completed replies. - **Agent Task Checklists**: Progress bars, recap for Claude Code, and reasoning effort for models. - **Git Branch and Worktree**: Track Git branches and worktrees. - **Model and Reasoning Effort**: Opt-in plan limits for 5-hour and weekly usage. - **Mascots and Animations**: Pixel-art mascots per tool, animated by their current activity. - **Smart Suppress**: No ping while you're already looking at the session tab. - **Glance Mode**: Silent and quiet hours, global shortcuts, and adjustable open/close speed. - **Silence Rules**: Hides in full screen, steps around menu-bar icons, and mutes while locked or asleep. - **Follow-Up Reminders**: Opt-in reminders for approvals, questions, and unfinished work. ### Tech Stack: Swift ### Why It Matters: CodeIsland provides a live status and approval system for AI coding tools, allowing developers to stay informed and make quick decisions without leaving their current window. This enhances productivity and collaboration, making it a valuable tool for developers working on multiple projects.

Overview: Chemprop is a repository for message passing neural networks designed for predicting molecular properties. It includes a ground-up rewrite and major release (v2.0.0) with a new major release (v2.0.0). The repository is free to use under the MIT License and has been cited in several papers. Key features: - Message Passing Neural Networks for Molecular Property Prediction - Ground-up rewrite and major release (v2.0.0) - New major release (v2.0.0) - Documentation available [here](https://chemprop.readthedocs.io/en/main/) - Tutorial notebooks in the `examples/` directory Tech stack: Python Why it matters: Chemprop is a valuable tool for chemists and chemical engineers looking to predict molecular properties using machine learning techniques. Its ground-up rewrite and new major release make it an efficient and modular package for chemical property prediction, making it a useful resource for researchers and practitioners in the field.

Overview: Lanhu MCP Server 提供了一个基于 Model Context Protocol (MCP) 的团队协作服务器，旨在提升 AI 编程团队的效率，自动分析需求并编写前后端代码，同时支持多种 AI 语言模型，如 Claude、Windsurf、OpenClaw 等。 Key features: - 自动分析需求并编写前后端代码 - 支持多种 AI 语言模型，如 Claude、Windsurf、OpenClaw - 无缝集成蓝湖（Lanhu）系统，提供设计稿和需求文档的自动下载和处理功能 - 本地 stdio 增加 `socksio` 与可选直连开关 - 免 Docker 本地启动指南 - 需求页支持按文字、稳定块 ID、分块 ID 或坐标精确取局部图 - 长页分片保持横向完整 - 系统架构：基于 Model Context Protocol (MCP) 的服务器 - 项目结构：包含核心功能、设计上下文、资源交付、性能优化等模块 - 高级配置：支持多种配置选项和性能指标监控 - 常见问题：提供常见问题解答和安全说明 - 贡献指南：鼓励贡献者提交代码和测试用例 - 许可证：MIT 协议 Tech stack: Python Why it matters: Lanhu MCP Server 通过自动分析需求并编写前后端代码，显著提高了 AI 编程团队的效率，

Overview: This repository contains the exercises and solutions from the book "An Introduction to Statistical Learning" in Python, addressing the challenge of implementing exercises in R for those using Python. Key features: - Exercises and solutions from "An Introduction to Statistical Learning" book - Python implementation of exercises - Solutions provided for conceptual questions - Commentary on notebooks for accuracy and research effort - Feedback and suggestions for improvement Tech stack: Python Why it matters: This repository offers a valuable resource for Python users who want to learn statistical learning concepts and intuition, making it an essential tool for data scientists and researchers.

### Overview: The Deeplearning4j Examples repository contains a collection of projects designed to support the development of deep learning applications using the Deeplearning4j ecosystem. This includes projects for building and training neural networks, data preprocessing, and various libraries for linear algebra and automatic differentiation. ### Key features: - **DL4J**: High-level API for building MultiLayerNetworks and ComputationGraphs, including support for Keras models and distributed training on Apache Spark. - **ND4J**: A general-purpose linear algebra library with over 500 mathematical and deep learning operations. - **SameDiff**: An automatic differentiation framework that supports TensorFlow frozen models and is planned for eager graph execution. - **DataVec**: A data processing tool for various machine learning formats and files. - **LibND4J**: A C++ library that underpins the JVM's access to native arrays and operations. ### Tech stack: Java ### Why it matters: The Deeplearning4j ecosystem provides a comprehensive set of tools for building and training deep learning models, making it a valuable resource for developers looking to implement and optimize deep learning applications.

### Overview: This repository contains summaries of various machine learning papers, focusing on advancements in unsupervised learning, object detection, audio processing, uncertainty, normalization, architectures, and more. It serves as a valuable resource for developers interested in staying updated with the latest developments in the field. ### Key Features: - **Unsupervised Learning**: Deep Clustering for Unsupervised Learning of Visual Features (ECCV 2018) - **Object Detection**: Deep Continuous Fusion for Multi-Sensor 3D Object Detection (ECCV 2018) - **Audio**: Audio-Visual Scene Analysis with Self-Supervised Multisensory Features (ECCV 2018) - **Uncertainty**: Towards Realistic Predictors (ECCV 2018) - **Object Detection**: Acquisition of Localization Confidence for Accurate Object Detection (ECCV 2018) - **Object Detection**: CornerNet: Detecting Objects as Paired Keypoints (ECCV 2018) - **Normalization**: Group Normalization (ECCV 2018) - **Architectures**: Spatial Transformer Networks (STN) (thanks, [alexobednikov](https://github.com/alexobednikov)) - **Loss Functions**: Working hard to know your neighbor’s margins: Local descriptor learning loss (thanks, [alexobednikov](https://github.com/alexobednikov)) - **Face Recognition**: Neural Aggregation Network for Video Face Recognition (

Overview: HuixiangDou is a professional knowledge assistant powered by LLM, designed for overcoming group chat scenarios and providing real-time streaming chat capabilities. It offers a complete suite of Web, Android, and pipeline source code, making it industrial-grade and commercially viable. Key features: - Advanced three-stage pipelines for group chat and real-time streaming chat. - No training required, with CPU-only and 2G, 10G configuration. - Web, Android, and pipeline source code available. Tech stack: Python. Why it matters: HuixiangDou addresses common group chat challenges by providing seamless and efficient knowledge assistance, making it a valuable tool for professionals and developers alike. Repository: InternLM/HuixiangDou GitHub stars: 2502

### Overview: Argo is a CNCF graduated project focused on simplifying the deployment and management of applications on Kubernetes. It includes tools like Argo Workflows, Argo CD, Argo Rollouts, and Argo Events, along with community contributions from [argoproj-labs]. The project aims to make complex Kubernetes workflows more accessible and efficient. ### Key Features: - **Argo Workflows**: A Kubernetes-native workflow engine for defining and executing complex tasks. - **Argo CD**: A declarative continuous delivery platform with a user-friendly UI. - **Argo Rollouts**: Simplifies Kubernetes deployment strategies, including Canary and Blue-Green deployments. - **Argo Events**: Event-based dependency management for Kubernetes. ### Tech Stack: - Kubernetes ### Why It Matters: Argo simplifies the deployment and management of applications on Kubernetes, making it easier for developers to scale and manage their workloads efficiently. Its ecosystem of tools and community contributions make it a valuable resource for anyone working with Kubernetes and continuous delivery.

Overview: 这是一本开源的NLP入门书籍，旨在为开发者提供从基础知识到实际应用案例的全面指导。 Key features: - 常用数据集简介 - 推荐更新源 - 打造NLP工具箱 - 如何衡量机器学习分类模型 - 词袋模型和TFIDF模型 - Word2Vec模型和Doc2Vec模型 - 自己动手训练word2vec模型 - 使用多层感知机进行文档分类 - 使用fasttext进行文档分类 - 使用LDA进行文档主题建模 - 使用Jieba进行中文词性标注 - 使用TextRank和TFIDF进行关键字自动提取 - 文档相似度 Tech stack: Python Why it matters: 本书通过开源的思想编写，为开发者提供了从基础到应用的全面指导，是学习NLP技术的绝佳资源。 Repository: duoergun0729/nlp Primary language: Python GitHub stars: 2496 GitHub description: 兜哥出品 一本开源的NLP入门书籍 README excerpt: # NLP基础知识 - [常用数据集简介](常用数据集简介.md) - [推荐更新源](推荐更新源.md) - [打造NLP工具箱](打造NLP工具箱.md) - [如何衡量机器学习分类模型](如何衡量机器学习分类模型.md) - [词袋模型和TFIDF模型](词袋模型和TFIDF模型.md) - [Word

Overview: SimCLR is a simple yet effective framework for contrastive learning of visual representations, implemented in PyTorch. Key features: - SimCLR is a PyTorch implementation of a simple contrastive learning method. - It is designed for unsupervised learning of visual representations. - The framework is simple to use and requires minimal configuration. Tech stack: PyTorch Why it matters: SimCLR provides a straightforward approach to contrastive learning, making it accessible for researchers and practitioners looking to improve their visual representation models without the need for complex data labeling.