Project directory

Projects

Open-source AI projects tracked by the leaderboard, ordered by GitHub stars.

Unknown 923 stars

mbreuss/diffusion-literature-for-robotics

Summary of key papers and blogs about diffusion models to learn about the topic. Detailed list of all published diffusion robotics papers.

### Overview: This repository provides a concise summary of key papers and blogs related to diffusion models, focusing on their application in robotics. It includes a categorized list of diffusion robotics papers and useful code repositories for starting a diffusion robotics project. ### Key Features: - **Diffusion Models Overview**: Brief introduction to diffusion models, including high-level math and general ideas. - **Diffusion in Robotics**: Detailed categorization of diffusion papers in robotics, covering topics such as imitation learning, video diffusion, online and offline reinforcement learning, inverse reinforcement learning, world models, task and motion planning, tactile sensing, and pose estimation. - **Code Implementations**: A list of code repositories for implementing diffusion models in robotics. ### Tech Stack: - Python ### Why It Matters: - Provides a comprehensive overview of diffusion models and their applications in robotics, making it a valuable resource for researchers and developers looking to understand and implement diffusion models in their robotics projects.

Kotlin 923 stars

woheller69/whoBIRD

Identify bird sounds in real time with this Android version of BirdNET. Bird sound recognition for more than 6,000 species worldwide.

Overview: whoBIRD is an Android app designed to identify birds by their sounds, providing accurate results without requiring an internet connection. It features an extensive database of over 6,000 bird species worldwide and uses advanced machine learning algorithms for real-time identification. Key features: - Real-time bird sound recognition - Extensive database of over 6,000 bird species - No internet connection required for operation - Automatic backup and deletion of database - Customizable options for detection notifications, image display, audio source, high pass filter, threshold, and language settings Tech stack: Kotlin Why it matters: whoBIRD offers a convenient and reliable way to identify birds by their sounds, making it essential for birders and nature enthusiasts worldwide. Repository: woheller69/whoBIRD GitHub stars: 923 GitHub description: Identify bird sounds in real time with this Android version of BirdNET. Bird sound recognition for more than 6,000 species worldwide. [Preview Image](https://www.youtube.com/embed/YYML_-e3yls)

Python 923 stars

zamalali/DeepGit

Deep research agent to help you find the best GitHub repositories 🕵️!

Overview: DeepGit is a powerful research tool designed to find the best GitHub repositories, preserving the full intent of your request and using semantic recall to surface hidden gems. Key features: - Preserves the full intent of your request - Uses semantic recall to improve the accuracy of the results - Judges every candidate based on real evidence - Reads the actual source of close calls before answering Tech stack: Python Why it matters: DeepGit helps developers find the repositories they actually meant, including hidden gems that a plain search would miss. Repository: zamalali/DeepGit GitHub stars: 922

Markdown 923 stars

zenstory-ai/novel-to-game

Agent skills for source-grounded novel-to-game adaptation (小说改编游戏 / story to game): cited design docs, target-runtime builds, evidence-based QA. Three playable browser examples. For Claude Code, Codex, and Kimi Code.

### Overview: NovelToGame is a set of seven open-source skills designed for the coding agent Claude Code, Codex, or Kimi Code. It adapts novels into playable browser games, focusing on source-grounded adaptation, target-runtime builds, and evidence-based QA. ### Key Features: - **Source-Grounded Adaptation**: Reads the entire novel and adapts it into a game direction. - **Target-Runtime Builds**: Designs the game world and builds it for runtime, ensuring the game runs as intended. - **Evidence-Based QA**: Validates the build through a recorded playthrough, ensuring the game is playable and meets the requirements. ### Tech Stack: Markdown ### Why It Matters: NovelToGame provides a practical and evidence-based approach to adapting novels into games, ensuring the quality and playability of the final product.

JavaScript 922 stars

Anil-matcha/open-spaces

Free, open-source alternative to ChatGPT Spaces: a self-hostable AI workspace for shared pages and conversations.

Overview: OpenSpaces is a free, open-source alternative to ChatGPT Spaces, providing a self-hostable workspace for shared pages, conversations, and AI tools. Key features: - Self-hostable infrastructure - Supports editable pages, conversations, and AI tools - Integrates with MuAPI for chat and image generation - Built to adapt and extend for personal needs Tech stack: Next.js, React, Tailwind CSS, FastAPI, SQLAlchemy, PostgreSQL-compatible database Why it matters: OpenSpaces offers a free, open-source alternative to ChatGPT Spaces, providing a self-hostable workspace for shared work, pages, and AI tools. It allows users to customize the application and integrate with their preferred AI services, making it a valuable tool for developers and creatives alike.

Shell 922 stars

Dicklesworthstone/claude_code_agent_farm

Orchestration framework for running 20+ Claude Code agents in parallel: automated bug fixing, best-practices sweeps, lock-based coordination, and real-time tmux monitoring

### Overview: Claude Code Agent Farm is a powerful orchestration framework that runs multiple Claude Code agents in parallel, improving your codebase through automated bug fixing, best practices implementation, and advanced coordination. ### Key Features: - **Parallel Processing**: Up to 50 Claude Code agents can run simultaneously. - **Multiple Workflows**: Supports bug fixing, best practices implementation, and coordinated multi-agent development. - **Agent Coordination**: Uses a lock-based system to prevent conflicts between parallel agents. - **Multi-Stack Support**: Compatible with 34 technology stacks including Next.js, Python, Rust, Go, Java, Angular, Flutter, C++, and more. - **Smart Monitoring**: Real-time dashboard with context warnings, heartbeat tracking, and tmux pane titles. - **Auto-Recovery**: Automatically restarts agents when needed with adaptive idle timeout based on work patterns. - **Progress Tracking**: Git commits with rich diff summaries and comprehensive HTML run reports. - **Context Management**: Agents automatically clear their own context when nearing limits, plus one-key broadcast of /clear to all agents (Ctrl+R). - **Highly Configurable**: JSON configs with variable substitution and dynamic chunk sizing. - **Flexible Viewing**: Multiple tmux viewing modes with shell completion support. - **Safe Operation**: Automatic settings backup/restore with size-based rotation, file locking, atomic operations. - **Development Setup**: 24 integrated tool installation scripts and pre-flight verification. - **Smart Controls**: Graceful shutdown with force-kill on double Ctrl+C within 3 seconds. ### Tech Stack: Python 3

Python 922 stars

mindscale-noah/MindMemOS

### Overview: MindMemOS is a memory operating layer designed for AI agents, enabling them to accurately remember user and task context, continuously improve memory quality, and integrate with various agents seamlessly. ### Key Features: - **Memory Persistence**: Stores user profiles, preferences, project facts, tool experience, and skill candidates across agents. - **Self-Evolving Memory**: Continuously improves memory quality through schema learning, dreaming, and feedback. - **Skill Integration**: Memories can be distilled into skill candidates, and skill execution results, failure traces, and user feedback are integrated back into the memory system. ### Tech Stack: Python ### Why It Matters: MindMemOS is a crucial component for AI agents, enabling them to efficiently remember and reuse context, continuously improve their memory quality, and integrate seamlessly with various agents, making them more capable and adaptable.

Java 921 stars

lionsoul2014/jcseg

Jcseg is a light weight NLP framework developed with Java. Provide CJK and English segmentation based on MMSEG algorithm, With also keywords extraction, key sentence extraction, summary extraction implemented based on TEXTRANK algorithm. Jcseg had a build-in http server and search modules for lucene,solr,elasticsearch,opensearch

### Overview: Jcseg 是一个基于 mmseg 算法的轻量级中文分词器，集成了关键字提取、关键短语提取、关键句子提取和文章自动摘要等功能。它还提供了一个基于 Jetty 的 web 服务器，支持多种搜索引擎接口，如 Lucene、Solr、Elasticsearch 和 OpenSearch。 ### Key Features: - **中文分词**: 采用 mmseg 算法和 Jcseg 独创的优化算法，支持七种切分模式。 - **关键字提取**: 基于 textRank 算法。 - **关键短语提取**: 基于 textRank 算法。 - **关键句子提取**: 基于 textRank 算法。 - **文章自动摘要**: 基于 BM25 + textRank 算法。 - **自动词性标注**: 采用词库 + 统计歧义去除计划，效果一般，不建议用于词性标注。 - **命名实体标注**: 采用词库 + 统计歧义去除计划，支持多种实体识别。 - **Restful API**: 嵌入 Jetty 提供高性能的 server 模块，支持多种语言客户端调用。 - **词库管理**: 支持自定义词库、词库多目录加载、词库分类、词库合并、词库更新自动加载等。 ### Tech Stack: Java ### Why It Matters: Jcseg 是一个功能丰富且性能

Python 921 stars

lonePatient/Bert-Multi-Label-Text-Classification

This repo contains a PyTorch implementation of a pretrained BERT model for multi-label text classification.

### Overview: This repository contains a PyTorch implementation of a pretrained BERT model for multi-label text classification. ### Key features: - Pretrained BERT and XLNET models for multi-label text classification. - PyTorch framework for efficient model training and inference. - Comprehensive structure with clear directory layout and dependencies. - Detailed README and GitHub description for easy understanding and usage. ### Tech stack: PyTorch ### Why it matters: This implementation provides a practical and efficient solution for multi-label text classification tasks, leveraging the strengths of BERT and XLNET models in a PyTorch environment.

Jupyter Notebook 920 stars

AntonioTepsich/Convolutional-KANs

This project extends the idea of the innovative architecture of Kolmogorov-Arnold Networks (KAN) to the Convolutional Layers, changing the classic linear transformation of the convolution to learnable non linear activations in each pixel.

### Overview: This project extends the innovative architecture of Kolmogorov-Arnold Networks (KAN) to Convolutional Layers, introducing learnable non-linear activations in each pixel. The resulting architecture, Convolutional Kolmogorov-Arnold Networks (CKAN), offers a more parameter-efficient alternative to traditional Multi-Layer Perceptrons (MLPs). ### Key Features: - **CKAN Architecture**: Extends KAN to Convolutional Layers, incorporating learnable non-linear activations. - **Efficient Implementation**: Utilizes an efficient implementation of KAN, based on the original implementation by KindXiaoming. - **Mathematical Foundation**: Based on the Kolmogorov-Arnold representation theorem, offering strong mathematical foundations. - **Parameter Efficiency**: Offers more parameter efficiency than traditional MLPs, with a reduced number of parameters compared to common convolutions. - **Experiments**: Includes experiments comparing CKAN with KAN, MLP, and classic convolutions, showcasing improved performance. ### Tech Stack: - **GitHub Repository**: AntonioTepsich/Convolutional-KANs - **Primary Language**: Jupyter Notebook ### Why It Matters: CKAN provides a novel and efficient alternative to traditional MLPs, offering better parameter efficiency and improved performance in various tasks.

Python 920 stars

candle-org/MindAct

MindSpore + 🤗Huggingface: Run any Transformers/Diffusers model on MindSpore with seamless compatibility and acceleration.

Overview: MindAct is a comprehensive toolkit for imitation learning policies, integrating datasets, policies, and environments with seamless compatibility and acceleration using MindSpore and Hugging Face. Key features: - Protocol-based architecture - Reproducible experiments - Optional dependencies via extras - Artifact management - Type-safe design Tech stack: MindSpore and Hugging Face Why it matters: MindAct simplifies the process of training and evaluating imitation learning policies, making it easier to reproduce results and ensure standardized evaluation protocols. Repository: candle-org/MindAct GitHub stars: 920 Primary language: Python GitHub description: MindSpore + 🤗Huggingface: Run any Transformers/Diffusers model on MindSpore with seamless compatibility and acceleration.

Python 919 stars

LibreYOLO/libreyolo

LibreYOLO is a MIT licensed open source computer vision library

Overview: LibreYOLO is a comprehensive computer vision library that provides a unified API for various tasks such as detection, segmentation, pose estimation, and more, all using the YOLOv9 model. It includes training and inference capabilities, making it easy to integrate into existing workflows with minimal changes. Key features: - Supports YOLOv9 and other core detectors - Includes training and inference capabilities - Compatible with a variety of export backends - Supports multiple models and training options - Provides speed optimization with features like fast-eval and hub-kernels - Includes optional extras for serving, models, training, and more Tech stack: Python Why it matters: LibreYOLO simplifies the process of integrating computer vision into projects by providing a single API for a wide range of tasks, making it easier to adapt existing workflows without extensive changes. Its modular design allows for easy upgrades and additions, making it a valuable tool for developers working in the field of computer vision.

Python 919 stars

MarvinTeichmann/KittiSeg

A Kitti Road Segmentation model implemented in tensorflow.

Overview: KittiSeg is a TensorFlow implementation of a Kitti Road Segmentation model that achieved first place on the Kitti Road Detection Benchmark. It is designed for real-time applications with impressive speed and state-of-the-art performance. Key features: - Semantic segmentation of roads using an FCN-based model - Achieved a MaxF1 score of over 96% on a small dataset of 250 images - Compatible with TensorFlow and TensorVision back end - Supports real-time inference and evaluation - Submodules for joint training of segmentation, classification, and detection Tech stack: Python Why it matters: KittiSeg is a valuable tool for researchers and developers working on autonomous driving and road segmentation tasks, providing a robust and efficient solution for real-world applications. Repository: MarvinTeichmann/KittiSeg

Python 919 stars

sberbank-ai-lab/LightAutoML

LAMA - automatic model creation framework

Overview: LightAutoML (LAMA) is an AutoML framework developed by Sber AI Lab, designed for automatic model creation across various tasks including binary classification, multiclass classification, and regression. It leverages the AutoWoE library for interpretable models and supports both independent and multitable datasets. Key features: - Automatic model creation for binary classification, multiclass classification, and regression tasks - Handles datasets with independent samples in each row - Supports multitable datasets and sequences (work in progress) - Uses AutoWoE library for interpretable models - Available on PyPI and GitHub Tech stack: Python Why it matters: LightAutoML simplifies the process of creating models, making it accessible for developers without extensive machine learning expertise. It automates the model creation process, reducing the time and effort required, and provides interpretable models for better understanding of the model's decisions. GitHub stars: 918 Repository: sberbank-ai-lab/LightAutoML

Ruby 919 stars

yoshoku/rumale

Rumale is a machine learning library in Ruby

Overview: Rumale is a machine learning library in Ruby that provides a Python-like interface for popular algorithms such as Support Vector Machines, Logistic Regression, Ridge, Lasso, Multi-layer Perceptron, Naive Bayes, Decision Tree, Gradient Tree Boosting, Random Forest, K-Means, Gaussian Mixture Model, DBSCAN, Spectral Clustering, Multidimensional Scaling, t-SNE, Fisher Discriminant Analysis, Neighbourhood Component Analysis, Principal Component Analysis, Non-negative Matrix Factorization, and many others. Key features: - Python-like interface for popular machine learning algorithms - Supports various algorithms including Support Vector Machines, Logistic Regression, Ridge, Lasso, Multi-layer Perceptron, Naive Bayes, Decision Tree, Gradient Tree Boosting, Random Forest, K-Means, Gaussian Mixture Model, DBSCAN, Spectral Clustering, Multidimensional Scaling, t-SNE, Fisher Discriminant Analysis, Neighbourhood Component Analysis, Principal Component Analysis, Non-negative Matrix Factorization, and many others - Uses Numo::NArray Alternative as a dependency for improved performance - Provides functions for loading and transforming datasets, as well as training and evaluating machine learning models Tech stack: Ruby Why it matters: Rumale simplifies the process of implementing machine learning algorithms in Ruby by providing a Python-like interface and supporting a wide range of algorithms. This makes it an attractive choice for developers looking to integrate machine learning capabilities into their Ruby projects.

Python 918 stars

EvelynFan/FaceFormer

[CVPR 2022] FaceFormer: Speech-Driven 3D Facial Animation with Transformers

### Overview: FaceFormer is a PyTorch implementation of the paper "FaceFormer: Speech-Driven 3D Facial Animation with Transformers" presented at CVPR 2022. It offers an end-to-end Transformer-based architecture for generating realistic 3D facial animations from speech inputs. ### Key Features: - **Transformer-based Architecture**: Utilizes Transformers for efficient and accurate 3D facial motion synthesis. - **Speech-Driven Animation**: Converts audio signals into 3D facial animations with accurate lip movements. - **Versatile Topologies**: Supports both BIWI and FLAME topologies for different applications. ### Tech Stack: - Python 3.7 - PyTorch 1.9.0 ### Why It Matters: FaceFormer provides a powerful tool for generating realistic 3D facial animations from speech inputs, making it valuable for applications in virtual reality, gaming, and entertainment. The use of Transformers allows for efficient and accurate synthesis, making it a significant advancement in the field of 3D facial animation.

Python 918 stars

LibreChat-AI/rag-api

ID-based RAG FastAPI: Integration with Langchain and PostgreSQL/pgvector

Overview: This project integrates Langchain with FastAPI in an Asynchronous, Scalable manner, providing a framework for document indexing and retrieval, using PostgreSQL/pgvector. Key features: - Document Management: Methods for adding, retrieving, and deleting documents. - Vector Store: Utilizes Langchain's vector store for efficient document retrieval. - Asynchronous Support: Offers async operations for enhanced performance. Tech stack: Python Why it matters: The ID approach allows for targeted queries when combined with file metadata stored in a database, such as is done by LibreChat, making it suitable for integration with LibreChat and other ID-based use cases.

Python 918 stars

google/langfun

OO for LLMs

### Overview: Langfun is a Python library designed to make language models (LMs) more accessible and easier to integrate into programming environments. It aims to simplify the interaction between natural language and computer programs by treating language as functions. ### Key Features: - **Object-Oriented Prompting**: Enables seamless integration between natural language and programming. - **Modular Prompts**: Facilitates a natural blend of texts and modalities. - **Efficient Workflow**: Suitable for both request-based workflows and batch jobs. - **Unified API**: Compatible with various popular LLMs like Gemini, GPT, Claude, and more. ### Tech Stack: Langfun is built on PyGlove, a Python library for building and managing models. ### Why It Matters: - **Powerful and Scalable**: Seamless integration, modular prompts, efficient workflows, and unified APIs. - **Simple and Elegant**: Intuitive programming model, easy integration, and comprehensive support for various LLMs. - **Agile Development**: Intellisense, easy debugging, minimal overhead, and support for agile development. ### Installation: Langfun can be installed using the following command: ```bash pip install langfun[all] ``` This installs all available extras, ensuring the minimal installation required. ### Hello, Langfun: ```python import langfun as lf import pyglove as pg from IPython import display class Item(pg.Object): name: str color: str class ImageDescription(pg.Object): items: list[Item] image = lf.Image.from

Python 918 stars

worldbench/DiffusionOPSD

🔥 On-Policy Self-Distillation in Diffusion Models

### Overview: DiffusionOPSD is an algorithm that turns image-level reward guidance into explicit targets for clean-output predictions at sampled queries. This approach is particularly useful for improving the quality of generated images in diffusion models. ### Key Features: - **Reward-guided Diffusion**: The algorithm uses reward guidance to refine the diffusion process, ensuring that the generated images are closer to the desired quality. - **Explicit Targets**: It constructs explicit targets for the diffusion model, which are directly related to the reward, making the process more efficient and effective. - **Efficient Training**: The algorithm reduces training time by approximately 30.5% through a more efficient reward-ascent step, broader prompt coverage, and stopping rollouts at the query. ### Tech Stack: Python ### Why It Matters: By efficiently refining the diffusion process with explicit targets, DiffusionOPSD significantly improves the quality of generated images in diffusion models, making them more aligned with the desired output. This approach not only enhances the performance of the model but also reduces the computational cost of training, making it a valuable addition to the field of diffusion models.

Python 917 stars

SkywalkerDarren/chatWeb

ChatWeb can crawl web pages, read PDF, DOCX, TXT, and extract the main content, then answer your questions based on the content, or summarize the key points.

Overview: ChatWeb is a versatile tool that can crawl web pages, extract text from PDF, DOCX, TXT files, and generate summaries. It utilizes GPT-3.5's embedding API and vector database to answer questions based on the content of the text, making it a valuable resource for developers and content creators. Key features: - Crawling web pages - Extracting text from PDF, DOCX, and TXT files - Generating summaries - Answering questions based on extracted content Tech stack: Python Why it matters: ChatWeb simplifies the process of extracting relevant information from large amounts of text and answering questions based on that content, making it an efficient solution for developers and content creators who often face token limits and need to break through them. Repository: SkywalkerDarren/chatWeb GitHub stars: 917 GitHub description: ChatWeb can crawl web pages, extract text from PDF, DOCX, TXT files, and generate an embedded summary. It can also answer your questions based on the content of the text. It is implemented using the chatAPI and embeddingAPI based on gpt3.5, as well as a vector database.