Research & Education

The PaddleOCR-VL-1.5 Online Demo provides a powerful platform for optical character recognition and visual language understanding. Users can easily upload an image or provide a URL, then select specific elements they wish to recognize, including plain text, complex tables, mathematical formulas, data-rich charts, or official seals. This tool is designed to showcase the capabilities of the PaddleOCR-VL-1.5 model, making advanced image analysis accessible for various applications. Hosted on Hugging Face, it offers a straightforward interface for testing and demonstrating the model's versatility in handling diverse visual recognition tasks.

awesome-3d-reconstruction-papers is a comprehensive collection of academic papers focused on 3D reconstruction within the deep learning era. This GitHub repository serves as a valuable resource for researchers and engineers, categorizing papers to facilitate easy navigation and discovery. The collection is organized into key areas such as object-level (single-view, multi-view, unsupervised), scene-level (single-view, multi-view), neural-surface, point-cloud, and RGB-D reconstruction, along with a survey section. Each entry typically includes the paper's representation, publisher, and links to project pages or code when available, making it an essential tool for staying updated on the latest advancements and methodologies in the field.

Awesome-Embodied-AI is a curated, open-source repository on GitHub that compiles an extensive list of papers and resources related to Embodied AI. Inspired by awesome-computer-vision, this project aims to continuously track and summarize the latest research and industrial advancements in the field. It includes sections on workshops, tutorials, talks, blogs, and various papers, categorized by surveys, Embodied AI and Robotics, Navigation, R&D, and LLM-Driven trends. The repository encourages community contributions through pull requests, making it a dynamic and collaborative resource for researchers and enthusiasts alike.

Trickle is an innovative AI agentic canvas designed to transform ideas into functional, live applications and websites. It offers a comprehensive suite of tools including built-in image and video generation, AI models, a database, and design capabilities, enabling users to build and launch production-ready apps and websites efficiently. This platform is ideal for individuals and teams looking to create web apps, AI-powered forms, landing pages, internal tools, multi-page apps, and interactive experiences without extensive coding knowledge. Trickle aims to make high-quality creation more accessible, faster, and significantly more affordable, allowing users to go from concept to product in hours rather than weeks.

AskVideo.ai is a powerful AI tool designed to transform how users interact with YouTube videos. It enables users to ask any question about a video and receive instant, accurate answers, complete with timestamps for easy reference. This eliminates the need to scrub through hours of content to find specific information. The platform supports various use cases, including academic learning, tutorials, business insights, research, and team collaboration. Users can paste any YouTube URL, and the AI processes the video by transcribing and indexing its content. AskVideo.ai also offers a command-line interface for developers, allowing chat interaction with YouTube videos directly from the terminal. It aims to make video learning more efficient and engaging for students, professionals, and content creators alike.

PracticeInterview.ai is an AI-powered platform designed to help job seekers hone their interview skills. It offers a comprehensive environment for practicing with real-world questions and scenarios, simulating actual interview experiences. Users receive instant feedback and personalized suggestions, enabling them to identify areas for improvement and refine their responses. This platform is ideal for anyone looking to boost their confidence and performance in job interviews, providing a practical and effective way to prepare for various professional roles and industries. By focusing on practical application and constructive criticism, PracticeInterview.ai aims to equip users with the necessary tools to succeed in their job search.

Essay Grader AI is an innovative AI tool designed to significantly reduce the time educators spend grading essays. By leveraging artificial intelligence, it can grade essays in seconds, providing teachers with immediate, high-quality, and specific feedback. This allows educators to focus more on teaching and less on administrative tasks. The platform aims to support thousands of teachers by streamlining the grading process, ensuring consistent and objective evaluations for student essays. It's a valuable resource for any educator looking to enhance efficiency and provide timely, constructive feedback to their students.

Microsoft Reading Coach is a free AI-powered reading practice tool designed to help individuals build their literacy skills. It leverages artificial intelligence to generate engaging stories and offers a library of leveled passages from ReadWorks, ensuring content matches learners' abilities and interests. The tool keeps users motivated by allowing them to unlock new characters and settings, practice their most challenging words, and monitor their progress over time. It is particularly beneficial for learners who already know how to decode words, providing a supportive environment for improving reading fluency and comprehension.

Spiking-Neural-Network offers a pure Python implementation of hardware-efficient spiking neural networks (SNNs). This tool focuses on developing a network capable of on-chip learning and prediction, utilizing modified learning and prediction rules that are energy-efficient and realizable on hardware. It incorporates the Spike-Time Dependent Plasticity (STDP) algorithm for network training, a biological process that modifies neural connections based on spike timing. The simulator supports classification tasks, employing a 'winner-takes-all' strategy for distinguishable results. Key features include neuron, synapse, receptive field, and spike train elements, along with functionalities for multi-class classification, variable threshold normalization, and lateral inhibition. The project also explores the generative property of SNNs to visualize learned patterns and discusses critical parameters like learning rate and weight initialization.

starVLA is an open-source research platform designed to facilitate the development of vision-language-action (VLA) models for generalist robots. It features a modular, 'Lego-like' codebase where functional components like models, data, trainers, and configurations follow a top-down, intuitive separation with high cohesion and low coupling. This design enables plug-and-play integration, rapid prototyping, and independent debugging. The framework supports various VLA architectures, including StarVLA-FAST, StarVLA-OFT, StarVLA-PI, and StarVLA-GR00T, and offers diverse training recipes such as supervised fine-tuning, multimodal co-training, and reinforcement learning adaptation. It integrates with broad benchmarks like LIBERO, RoboCasa, and Calvin, and provides a model zoo with released checkpoints.

StudyRaid is an AI-powered platform designed to revolutionize learning by generating comprehensive, personalized courses on a vast array of subjects. Users can instantly create courses complete with engaging lessons, interactive quizzes, smart flashcards, and simulated exams. The platform offers two AI models: Ultra for complex subjects with audio lessons and near-perfect accuracy, and Light for faster generation of simpler topics. StudyRaid aims to adapt to individual learning styles, providing certifications to validate skills and offering real-time AI chat for questions. It's accessible across web, iOS, and Android devices, making learning flexible and efficient for students and educators alike.

T2M-GPT is an open-source PyTorch implementation for generating human motion from textual descriptions, as detailed in its CVPR 2023 paper. The tool utilizes discrete representations to create realistic motion sequences. It includes functionalities for VQ-VAE and GPT training, evaluation, and SMPL mesh rendering. Users can install the environment, prepare datasets like HumanML3D and KIT-ML, and download pre-trained models and motion/text feature extractors. A quick start guide is available via a Jupyter Notebook demo, and the project offers visual results, installation instructions, and detailed steps for training and evaluating both VQ-VAE and GPT models. The project also provides a HuggingFace space demo for both skeleton and SMPL mesh visualization.

Street-fighter-ai is an AI agent specifically designed and trained using deep reinforcement learning to play the classic game "Street Fighter II: Special Champion Edition." The agent operates by making decisions based solely on the RGB pixel values of the game screen, demonstrating a sophisticated approach to game AI. It has been shown to achieve a 100% win rate in the first round of the final level, though this can involve overfitting. The project provides detailed instructions for environment setup, running tests with pre-trained models, and even training your own models. It leverages open-source libraries like OpenAI Gym Retro and Stable-Baselines3, making it a valuable resource for researchers and enthusiasts in AI and reinforcement learning.

T-Rex2 is an advanced object detection model developed by IDEA-Research, designed to overcome the limitations of traditional, closed-set object detection systems. By integrating both text and visual prompts, T-Rex2 harnesses the strengths of both modalities, providing robust zero-shot capabilities. This makes it a versatile tool for identifying and locating objects within images across a wide range of applications, including agriculture, industry, livestock monitoring, biology, medicine, OCR, retail, electronics, transportation, and logistics. It supports three main workflows: interactive visual prompt, generic visual prompt, and text prompt, covering most object detection scenarios. The project provides API access and a local Gradio demo for easy implementation and experimentation.

TextRank is a Python implementation of the TextRank algorithm, specifically designed for automatic keyword and sentence extraction, which facilitates summarization. This particular implementation distinguishes itself by utilizing Levenshtein distance to determine the relationship between text units, offering a unique approach to text analysis. The project is based on the foundational paper "TextRank: Bringing Order into Text" by Rada Mihalcea and Paul Tarau. It provides functionalities for both keyword and sentence extraction, making it a valuable tool for researchers and developers working with text data. The library is installable via pip and requires NLTK resources, which can be fetched using a simple command.

tf-cpn is a Tensorflow re-implementation of the Cascaded Pyramid Network (CPN), a state-of-the-art model for multi-person pose estimation that won the 2017 COCO Keypoints Challenge. This open-source tool provides researchers and developers with the code and pre-trained models necessary to implement and experiment with advanced pose estimation. It includes detailed instructions for training on the MSCOCO dataset, downloading base models, and running validation tests. The repository also offers pre-trained models for various configurations (ResNet-50, ResNet-101 with different input sizes) and provides performance metrics on COCO minival and test-dev datasets, making it a valuable resource for academic and practical applications in computer vision.

text_gcn is an open-source implementation of Graph Convolutional Networks (GCNs) specifically designed for text classification tasks. This tool provides the necessary code to reproduce the results presented in the paper "Graph Convolutional Networks for Text Classification" from the AAAI 2019 conference. It requires Python 2.7 or 3.6 and Tensorflow >= 1.4.0, making it accessible for those familiar with these environments. The repository includes scripts for data preparation, graph building, and model training, along with examples for various datasets like 20ng, R8, R52, ohsumed, and mr. An inductive version, fast_text_gcn, is also available for scenarios where test documents are not included in the training process.

ThoughtSource is an open and central resource designed for researchers and developers working with chain-of-thought reasoning in large language models. It provides a comprehensive collection of datasets, including general question answering, scientific/medical QA, and math word problems, all formatted for standardized chain-of-thought analysis. The platform also includes tools for generating reasoning chains with various language models (OpenAI, Hugging Face) and evaluating their performance. With its dataset annotator and viewer applications, ThoughtSource aims to foster a community around improving trustworthy and robust reasoning in AI, particularly for scientific research and medical practice. It is developed by the Samwald research group.

Knowt is a comprehensive AI study tool designed to be the #1 free alternative to Quizlet, trusted by millions of students and teachers. It offers a suite of AI-powered features including an AI Lecture Notetaker that records lectures and instantly generates detailed notes, flashcards, quizzes, and games. Users can also leverage its AI PDF Summarizer to transform long readings into study guides and interactive recall methods. Knowt provides unlimited free access to study modes like Learn, Match, and Practice Test, which are often paid features on other platforms. Additionally, it includes a Voice Tutoring & Podcasts feature, allowing Kai, the AI, to quiz users verbally or create podcasts from uploaded content. Knowt supports various file types, including PDFs, PowerPoints, lecture videos, and audio, making it a versatile study companion for any subject.

UNETR++ is an open-source tool designed for efficient and accurate 3D medical image segmentation, developed by researchers from Mohamed Bin Zayed University of Artificial Intelligence, University of California Merced, Google Research, and Linkoping University. It addresses the computational bottleneck of traditional self-attention mechanisms in volumetric medical imaging by introducing a novel efficient paired attention (EPA) block. This block efficiently learns spatial and channel-wise discriminative features with linear complexity, reducing parameters, compute cost, and inference speed. The tool has been extensively evaluated on five benchmarks, including Synapse, BTCV, ACDC, BRaTs, and Decathlon-Lung, demonstrating state-of-the-art performance with significant efficiency gains. It is available in Keras 3 as part of the AI Toolkit for Healthcare Imaging.

Unet-Segmentation-Pytorch-Nest-of-Unets is an open-source project offering a comprehensive collection of Unet model implementations for image segmentation tasks using PyTorch. This tool provides various architectures, including the original Unet, RCNN-Unet, Attention Unet, RCNN-Attention Unet, and Nested Unet (UNet++). It is designed for developers and researchers working on biomedical image segmentation or other image analysis problems. The repository includes code for data loading, model definitions, metrics, and visualization, making it a valuable resource for experimenting with and applying different Unet-based segmentation models. Users can easily clone the repository, install dependencies, and configure data paths to run the models.

Opennote, formerly known as Deep Tutor, is an AI-powered notebook designed to enhance learning and knowledge management. It allows users to capture and organize notes with a rich editor, and interact with Galileo, an AI thinking partner, to ask questions, explore ideas, and receive clear explanations. The platform can generate videos, diagrams, and visual explanations for concepts, create inline flashcards and quizzes, and record audio or process YouTube links for instant notes and summaries. Opennote aims to centralize learning materials, providing proactive suggestions and practice problems to help users understand and retain information more effectively.

MIND-Interview is an AI-powered platform designed to assist both job seekers and recruiters in the interview and hiring process. For job seekers, it offers AI-driven interview coaching to help them prepare effectively and create compelling video resumes. Recruiters can leverage the platform for resume analysis and to conduct AI-powered interviews, which aims to streamline candidate screening and evaluation. The tool focuses on enhancing the efficiency and effectiveness of recruitment by providing intelligent insights and automated functionalities, ultimately aiming to improve the matching process between candidates and roles.

MySivi is an AI English speaking tool designed to help users master English speaking online. It features Arya, an AI English teacher, who provides instant feedback on pronunciation, grammar, and fluency during real conversations. The platform supports scenario-based learning for various situations like job interviews or daily chats, and offers a personalized learning path for all skill levels. Users can also connect with co-learners globally for practice calls. MySivi tracks progress, offers multi-language support for learners, and covers a wide range of topics, making it a comprehensive solution for improving English speaking skills.