Research & Education

llm-twin-course is a free educational resource designed to guide users through the process of building a production-ready Large Language Model (LLM) and Retrieval Augmented Generation (RAG) system. The course emphasizes LLMOps best practices, offering practical, hands-on lessons and accompanying source code. It covers the entire development lifecycle, from initial data gathering to the final stages of productionizing LLMs, with a specific focus on creating an AI replica.

hands-on-rl is a comprehensive and free course aimed at individuals with no prior knowledge in reinforcement learning (RL). It guides users through the fundamentals and advanced concepts of RL, enabling them to gain proficiency in the field. The course emphasizes a practical, step-by-step methodology for both understanding and implementing various RL algorithms and theories. Being open-source and available on GitHub, it promotes collaborative learning and accessibility for a wide audience interested in artificial intelligence and machine learning.

Animatable_nerf is an open-source research tool that provides the implementation for "Animatable Implicit Neural Representations for Creating Realistic Avatars from Videos," a paper accepted to TPAMI 2024 and ICCV 2021. This tool allows researchers to generate realistic avatars from video footage by leveraging animatable neural fields. It supports various configurations, including vanilla Animatable NeRF, versions with neural blend weight fields replaced by displacement fields, and versions where the canonical NeRF model is replaced with a neural surface field (SDF output). The repository includes evaluation frameworks for reconstruction quality comparison and provides access to datasets like Mobile-Stage and SyntheticHuman++ for further research and development in neural rendering and 3D human body modeling.

Llama 3.1 70b Demo is an AI chatbot specifically designed for engaging in conversational tasks. Its core capabilities include advanced language understanding and efficient text generation. This tool can serve as a valuable educational resource, providing a platform for users to interact with and learn from an AI. It is offered to users at no cost.

Magiv2 Demo is an AI chatbot specifically designed to assist with automation and content generation tasks. It serves as a versatile tool for users looking to streamline various processes and create content efficiently. The chatbot is particularly useful for educational applications, providing a resource for learning and practical implementation. Additionally, it can handle general task automation, making it a valuable asset for individuals seeking to enhance productivity. The tool is accessible for free on the Hugging Face platform.

Speeko AI Speech Coach is designed to build public speaking skills, helping users become confident and effective speakers. The app tracks voice and speech patterns in real-time, providing simple, friendly alerts and personalized feedback on pace, eloquence, intonation, sentiment, talk time, and word choice. It offers bite-sized exercises tailored to unique speaking styles and goals, allowing users to practice and improve in just a few minutes a day. Speeko also includes a complete toolkit with digital notecards, interview prompts, and vocal warm-ups. Users can access exclusive content from America's #1 voice coach, Roger Love, and even work 1:1 with a coach. The tool is trusted by over 400,000 professionals and is available for Mac, iPhone, and iPad.

Qwen2 VL Localization is an AI tool designed for visual localization, enabling users to detect and pinpoint objects within images. By uploading an image and providing a descriptive text prompt, the application processes the input to identify the requested objects. The output includes a list of precise bounding boxes for each detected object, along with an annotated image that visually highlights these objects. This tool is particularly useful for tasks requiring detailed object identification and localization, making it suitable for AI research and development in image analysis. It is available for free on Hugging Face, offering an accessible solution for visual detection needs.

ssm is a powerful tool designed for Bayesian learning and inference within state space models. It offers comprehensive functionalities for simulating, learning, and performing inference across a variety of state space models. The project is currently undergoing a JAX refactor, which aims to leverage JIT compilation and provide enhanced support for GPU and TPU hardware, significantly boosting performance and computational efficiency for complex scientific computing tasks. This makes ssm particularly valuable for researchers and data scientists working with dynamic systems and requiring robust statistical modeling capabilities.

Deep3DFaceRecon_pytorch is an open-source PyTorch implementation for accurate 3D face reconstruction, building upon the original TensorFlow version. It utilizes weakly-supervised learning to reconstruct 3D faces from single images or image sets, offering improved accuracy and visual consistency. Key enhancements include a differentiable renderer using Nvdiffrast, Arcface for perceptual loss computation, and data augmentation during training. The tool achieves state-of-the-art performance on various datasets like FaceWarehouse, MICC Florence, and the NoW Challenge. It supports both inference with pre-trained models and training new models from scratch, making it suitable for researchers and developers in computer vision and 3D modeling.

sdfstudio is a unified, open-source framework designed for neural implicit surface reconstruction, leveraging the foundation of the Nerfstudio project. It provides a modular architecture that allows for the implementation and exploration of different surface reconstruction methods, such as UniSurf, VolSDF, and NeuS. The framework supports various scene representations and datasets, making it a versatile tool for advanced 3D modeling, research, and development in the field of neural implicit surfaces. Its open-source nature encourages community contributions and provides a flexible platform for experimenting with cutting-edge 3D reconstruction techniques.

AskIndra is a Wellness & Lifestyle tool designed to make environmental data accessible and actionable. It transforms real-time weather and air-quality information into clear, human-readable guidance, eliminating the need for users to interpret complex dashboards or raw data. Instead, users can ask simple questions about their environment and receive understandable advice, focusing on how conditions might impact their daily lives. This tool aims to provide decision-ready insights, making it easier for individuals to understand and respond to current environmental conditions.

RaDe-GS, or Rasterizing Depth in Gaussian Splatting, is a cutting-edge Content & Design tool developed by HKUST-SAIL. It significantly enhances the performance and accuracy of 3D scene reconstruction and rendering by incorporating advanced techniques like multi-view regularization and refined densification strategies. The project provides updated code and formulations, enabling users to achieve superior results on challenging datasets such as DTU and Tanks and Temples. It also supports novel view synthesis and geometry evaluation, making it a powerful resource for researchers and developers working with 3D Gaussian Splatting. The tool is built upon the original 3D Gaussian Splatting implementation and integrates ideas from several recent works to offer a robust and efficient solution for 3D graphics tasks.

R-FCN (Region-based Fully Convolutional Networks) is an open-source object detection framework designed for computer vision research and applications. It utilizes deep fully-convolutional networks to achieve accurate and efficient object detection. Unlike previous region-based detectors that apply costly per-region sub-networks, R-FCN shares almost all computation on the entire image, making it highly efficient. The framework can integrate powerful fully convolutional image classifier backbones, such as ResNets, for enhanced performance. It supports end-to-end training and inference for object detection and has been tested on Windows and Ubuntu platforms, requiring MATLAB and a Caffe build.

LuaRadio is a lightweight and embeddable flow graph signal processing framework specifically designed for software-defined radio (SDR). Built on LuaJIT, it offers a small binary footprint and no external hard dependencies, making it highly portable. The framework provides a comprehensive suite of source, sink, and processing blocks, along with a simple API for defining and running flow graphs, creating custom blocks, and managing data types. It's ideal for rapidly prototyping software radios, developing modulation/demodulation utilities, and conducting signal processing experiments. LuaRadio can also be embedded into existing radio applications, serving as a user-scriptable engine for advanced signal processing tasks. It supports computational acceleration through LuaJIT's FFI to wrap external libraries like VOLK, liquid-dsp, and others, ensuring efficient performance.

Trostun is a full-service digital marketing agency dedicated to elevating businesses into strong brands through innovation. They provide comprehensive solutions across branding, development, and digital marketing, aiming to improve recognition, outreach, and connectivity with customers. Their services include expert branding and identity recreation, UI/UX design focusing on aesthetics, navigation, and visuals, and smart web development to create impactful commercial websites. Trostun emphasizes a customer-centric approach, employing a multi-pronged strategy to deliver measurable outcomes and help businesses achieve their maximum potential in the digital landscape.

swift-embedded-examples is a collection of demonstration projects designed to help developers learn and implement Embedded Swift. This compilation and language mode allows for the development of baremetal, embedded, and standalone software using the Swift programming language. The repository serves as a valuable resource for understanding how to leverage Swift in embedded systems, offering practical examples that illustrate various functionalities and use cases. It aims to simplify the process of getting started with Embedded Swift development by providing ready-to-use code and project structures, making it easier for developers to explore and build their own embedded applications.

SoundMind is an innovative project that provides a rule-based reinforcement learning (RL) algorithm specifically designed to endow audio language models (ALMs) with deep bimodal reasoning abilities. It is built upon the Audio Logical Reasoning (ALR) dataset, which comprises 6,446 text-audio annotated samples tailored for complex reasoning tasks. This resource enables the training of ALMs to perform sophisticated logical reasoning across both audio and textual modalities. The repository offers the official implementation, dataset download links, environment setup instructions, and details for RL-training and evaluation, making it a valuable tool for researchers and developers in the field of audio-language processing.

FastGS is an acceleration framework designed to supercharge 3D Gaussian Splatting training, enabling state-of-the-art results within 100 seconds. This represents a substantial speed improvement, being 3.32 times faster than DashGaussian on the Mip-NeRF 360 dataset and offering a 15.45 times acceleration compared to vanilla 3DGS on Deep Blending. Despite its rapid training, FastGS maintains comparable rendering quality to other state-of-the-art methods. The framework is highly versatile, seamlessly integrating with various backbones like Vanilla 3DGS, Scaffold-GS, and Mip-splatting. It is also proven effective across multiple tasks, including dynamic scenes, surface reconstruction, sparse-view, large-scale, and SLAM tasks. FastGS is memory-efficient, requiring low GPU memory, and offers easy deployment with a simple post-training tool.

VideoSuperResolution is an open-source project offering a comprehensive collection of state-of-the-art video and single-image super-resolution architectures. These models are reimplemented in TensorFlow, with several referenced PyTorch implementations also included. The project provides a simple, easy-to-use framework for training and data processing based on TensorFlow, capable of handling raw NV12/YUV as well as sequences of images as inputs. Users can install the package via PyPI and download pre-trained weights for various models like SRCNN, VESPCN, and ESRGAN. It supports a wide range of datasets for training and testing, making it a valuable resource for researchers and developers working on image and video enhancement.

Latex Ocr is a specialized tool engineered to transform images containing mathematical formulas and equations directly into Latex code. This functionality is particularly beneficial for users who frequently work with academic or scientific documents. By enabling the extraction and digitization of complex mathematical expressions from visual sources, Latex Ocr streamlines the process of incorporating these elements into Latex-based projects. It serves as a valuable resource for individuals in educational and research fields.

SummarizeThis.io is a tool built to provide concise summaries of diverse content formats. Users can input text, web page URLs, PDF documents, or YouTube video links, and the tool will generate a condensed version of the information. This functionality is particularly useful for quickly grasping the main points of lengthy articles, reports, or video content without needing to consume the entire source. It aims to save time and improve information digestion.

Awesome-Tabular-LLMs provides a comprehensive, curated list of research papers specifically focused on the application of Large Language Models (LLMs) to various table-related tasks. This resource is designed to keep researchers and practitioners updated on the latest developments in the field. It covers a range of applications, including but not limited to, table question answering, where LLMs interpret and respond to queries based on tabular data; table-to-text generation, which involves converting structured table data into natural language descriptions; and text-to-SQL conversion, enabling users to generate SQL queries from natural language prompts. The primary goal is to serve as a valuable reference for anyone interested in the intersection of LLMs and tabular data processing.

ObjectDetection-OneStageDet is an open-source object detection framework developed by Tencent, designed to provide a unified platform for single-stage generic object detectors. Currently, it supports YOLOv2 and YOLOv3 implementations, with future plans to integrate YOLO and SSD into a single framework. The tool emphasizes performance and speed, offering good mAP scores and fast inference times, especially with various efficient backbones like TinyYOLO, MobileNet, and ShuffleNet. It provides comprehensive instructions for installation, data preparation, training, evaluation, and benchmarking, making it suitable for developers and researchers working on object detection tasks.

Mamba_State_Space_Model_Paper_List is an open-source, curated list of research papers focused on State-Space Models and Mamba. Mamba is presented as a new-generation network alternative to Transformers. This resource, maintained on GitHub, provides a comprehensive collection of papers that delve into both the theoretical foundations and practical applications of Mamba models. It is specifically designed to be a valuable asset for researchers and practitioners working within this specialized field.