Research & Education

The Video Generation Leaderboard is a Hugging Face Space designed to provide a comprehensive comparison of text-to-video and image-to-video generation tools. It serves as a valuable resource for users to evaluate the performance and capabilities of different AI models in the video generation domain. By offering a centralized platform, it helps researchers, developers, and enthusiasts stay informed about the latest advancements and identify the most effective tools for their specific needs. The leaderboard facilitates informed decision-making by presenting a clear overview of various services, making it easier to select the best AI video generation solution.

Whisper vs Distil-Whisper is an AI tool designed to facilitate the comparison between the original Whisper model and the Distil-Whisper model for audio transcription tasks. This platform allows users to evaluate the accuracy and speed of transcriptions generated by both models, providing insights into their respective performances. It serves as a valuable resource for developers and researchers interested in speech-to-text technologies, offering a direct way to benchmark and understand the differences between these two prominent AI models. The tool is hosted on Hugging Face Spaces, indicating its accessibility and community-driven nature.

WikiRacing Language Models is an interactive AI tool hosted on Hugging Face Spaces, designed for users to engage in a competitive quiz format against language models. This platform offers a unique opportunity to test one's knowledge and understanding in a fun, game-like setting. While the current live website indicates a runtime error, the tool's core concept revolves around pitting human intelligence against AI in a race to answer questions. It serves as an experimental ground for AI researchers and language model enthusiasts to observe and interact with AI capabilities in a practical, engaging scenario.

Weights2Weights Demo, hosted on Hugging Face Spaces by Snap Research, is an AI-powered tool designed for creating and editing images of custom identities. Users can either sample a new identity or upload an existing image to begin. The tool provides intuitive sliders to fine-tune various facial attributes, including youth, nose shape, hair, and eyebrows, offering a high degree of control over the generated output. This makes it suitable for exploring different appearances and generating unique portraits based on specific attribute adjustments. It's a practical demonstration of advanced image manipulation capabilities.

The YKS_2025_LLM_Leaderboard is a specialized platform designed for evaluating and comparing large language models (LLMs) against the challenging 2025 YKS university entrance exam. This tool provides a clear, ranked table showcasing various LLMs, detailing their overall performance through total points, and offering granular insights with subject-wise scores. It serves as a valuable resource for researchers, educators, and anyone interested in assessing the capabilities of AI models in an academic context. The leaderboard allows users to filter results by model name or score, facilitating easy navigation and comparison. Hosted on Hugging Face, it aims to contribute to AI research and educational understanding by providing a standardized benchmark.

XL Model Experiments is a free AI tool hosted on Hugging Face that enables users to generate high-quality images from text descriptions. Users can input a prompt, select from various presets, and fine-tune parameters such as image size, quality, and seed to achieve desired visual outcomes. This application is designed for experimenting with AI models, providing a user-friendly interface for exploring the capabilities of text-to-image generation. It's an accessible platform for both beginners and those with more experience in AI art, offering a straightforward way to create unique images based on textual inputs.

XTTS_V1 -> V2 work on CPU Can duplicate is a free AI voice generator tool hosted on Hugging Face, developed by Olivier-Truong. This application enables users to generate speech in various languages by providing a text prompt and a reference audio clip. Users have the flexibility to either upload an existing audio file or record a sample directly using their microphone. The tool is designed to facilitate experimentation with voice cloning and duplication on CPU, leveraging the capabilities of XTTS models. It's an accessible platform for those looking to explore speech synthesis without requiring high-end GPU resources.

Librar Labs provides an intelligent library system designed to streamline operations for school libraries, even in the absence of a dedicated librarian. Its AI-powered features include rapid search capabilities, smart recommendations for book acquisitions, and efficient bulk scanning for inventory management. The platform allows users to catalog new arrivals, identify misplaced items, and conduct inventory checks simply by pointing a phone at shelves. Librar Mobile enables full library management from a mobile device, supporting ISBN, barcodes, and RFIDs. It aims to transform book rooms into functional libraries, reducing manual handling and making library management accessible and efficient for schools globally.

BERT4doc-Classification is an open-source project offering code and resources specifically designed for fine-tuning BERT models for text classification tasks. It provides a comprehensive solution based on extensive experiments detailed in the paper "How to Fine-Tune BERT for Text Classification?". The project includes requirements for both further pre-training (using TensorFlow 1.1x) and fine-tuning (using PyTorch). Users can prepare various datasets, including Sogou News and others built by Zhang et al., and leverage Google BERT models. The repository guides users through generating pre-training corpora, running further pre-training, and fine-tuning on downstream tasks with detailed command-line examples. It also addresses considerations for different GPU setups and offers advanced fine-tuning arguments like layer-wise learning rates and strategies for handling long texts.

Rocking Robots is an independent news platform dedicated to exploring the intersection of humanity, robotics, and artificial intelligence. It provides in-depth backgrounds, reports, and news on how technology is transforming society, impacting individuals, businesses, governments, and healthcare sectors. The platform focuses on tracking the digital transformation and highlighting the key individuals and innovations shaping this new world. Content categories include Bots & Business, Bots & Brains, Bots in Society, Bots & Bullets, Bots in Beeld, and People in Robotics, offering a comprehensive view of the AI and robotics landscape.

FoodVision Mini is an AI-powered image classification tool hosted on Hugging Face Spaces. Users can upload an image of food, and the application will classify it into one of three categories: pizza, steak, or sushi. In addition to the classification, the tool also provides the prediction time, offering a quick and efficient way to categorize food items. This tool is suitable for anyone interested in basic food image recognition, particularly those exploring machine learning applications or needing quick food identification for simple tasks.

Instruct X-Decoder is an AI tool hosted on Hugging Face, designed for various code-related tasks. While its specific functionalities are currently unavailable due to a build error, the platform it resides on, Hugging Face, offers extensive resources for machine learning applications, including models, datasets, and spaces for hosting AI demos. The tool's presence on Hugging Face suggests a focus on automation and potentially content generation within a coding context, aligning with educational and development purposes. Users interested in code assistants and AI-driven development tools would typically explore such offerings.

LiveAvatar is an open-source implementation of the research paper "Live Avatar: Streaming Real-time Audio-Driven Avatar Generation with Infinite Length." This algorithm-system co-designed framework allows for real-time, streaming, and interactive avatar video generation of infinite length. Powered by a 14B-parameter diffusion model, it achieves 45 FPS on multi-card H800 GPUs with 4-step sampling and supports Block-wise Autoregressive processing for videos exceeding 10,000 seconds. Key highlights include real-time streaming interaction with low latency, infinite-length autoregressive generation, and strong generalization across cartoon characters, singing, and diverse scenarios. The project provides code for both multi-GPU and single-GPU inference, including a Gradio Web UI, and supports FP8 quantization for 48GB GPUs.

TokenFormer is the official implementation of the ICLR2025 Spotlight paper, "TokenFormer: Rethinking Transformer Scaling with Tokenized Model Parameters." This tool introduces a fully attention-based neural network that unifies token-token and token-parameter interactions, maximizing the flexibility of neural network architectures. By tokenizing both data and model parameters, TokenFormer inherently enhances model scalability, allowing for progressively efficient scaling. The architecture is designed to be natively scalable, leveraging attention mechanisms for interactions between input tokens, and between tokens and model parameters. This approach aims to offer greater flexibility than traditional Transformers, contributing to advancements in foundation models, sparse inference (MoE), parameter-efficient tuning, device-cloud collaboration, and vision-language applications.

Transformer-TTS is a PyTorch implementation of the "Neural Speech Synthesis with Transformer Network," designed for efficient and high-quality speech synthesis. This model boasts training speeds 3 to 4 times faster than well-known seq2seq models such as Tacotron, while maintaining comparable synthesized speech quality. It utilizes a post-network based on the CBHG model from Tacotron and converts spectrograms into raw audio waves using the Griffin-Lim algorithm. The project includes detailed instructions for data preparation, training the autoregressive attention network and post-network, and generating TTS samples, making it a valuable resource for researchers and developers in speech synthesis.

Deep-Photo-Enhancer is an open-source project offering a TensorFlow implementation of the CVPR 2018 spotlight paper, "Deep Photo Enhancer: Unpaired Learning for Image Enhancement from Photographs with GANs." This tool allows users to enhance photographs using deep learning, specifically Generative Adversarial Networks (GANs), without the need for paired input and output images for training. It includes models for both supervised and unsupervised learning, and provides a simplified tutorial for processing images. The project also highlights improvements like global U-Net, adaptive WGAN (A-WGAN), and individual batch normalization (iBN) for better results in various applications beyond photo enhancement.

DeepGBM is a deep learning framework specifically designed for online prediction tasks, leveraging the power of Gradient Boosting Decision Trees (GBDT) for distillation. Presented at KDD'2019, this framework aims to significantly improve prediction accuracy in real-time scenarios. It integrates GBDT-based models, specifically LightGBM, with PyTorch-based neural networks. The project includes comprehensive code for data preprocessing, baseline model implementations, and the proposed DeepGBM model. Users can prepare their data in CSV format, process it through encoders, and then load numerical and categorical data for training. The framework supports training GBDT2NN or the full DeepGBM model, offering flexibility for different prediction needs.

Difix3D is an open-source project designed to enhance 3D reconstructions by leveraging single-step diffusion models. It offers a comprehensive framework for improving the quality of 3D data, specifically targeting artifact removal and the refinement of novel views. The tool provides both Difix for single-step diffusion artifact removal and Difix3D for progressive 3D updates, including integration with popular 3D reconstruction frameworks like Nerfstudio and gsplat. Additionally, Difix3D+ introduces real-time post-rendering capabilities to further sharpen details and improve visual fidelity. This makes it a valuable resource for researchers and developers working on advanced 3D computer vision tasks, offering practical implementations and models for immediate use.

GenerativeImage2Text (GIT) is a repository from Microsoft that provides code examples and pre-trained models for generating text from images. It leverages a Generative Image-to-text Transformer for various vision and language tasks. Users can perform image captioning, where the model describes the content of an image, or visual question answering, where the model answers questions about an image. The tool supports inference on single images, multiple frames (for video analysis), and TSV files containing collections of images. It offers different model sizes (base and large) and fine-tuned versions for specific datasets like COCO, VQAv2, and TextCaps, allowing for tailored performance across diverse applications.

gpt_paper_assistant is an open-source, GPT-4 based tool designed to help researchers stay updated with the latest papers on ArXiv. It functions as a personalized daily scanner, identifying papers relevant to specified topics and authors. The tool leverages GPT-4 for evaluating paper relevance and novelty, and can filter papers based on author matches and semantic scholar IDs. It runs automatically via GitHub Actions, publishing daily summaries to a static GitHub Pages website or posting directly to Slack. Users can customize topics, authors, and filtering thresholds, making it a highly adaptable solution for academic research.

GraphWaveletNeuralNetwork is an open-source PyTorch implementation of the "Graph Wavelet Neural Network" (GWNN) as presented at ICLR 2019. This novel graph convolutional neural network addresses limitations of previous spectral graph CNN methods by utilizing graph wavelet transform, which avoids computationally expensive matrix eigendecomposition. The graph wavelets are sparse and localized, enhancing efficiency and interpretability for graph convolution tasks. The tool is designed for researchers and machine learning engineers working with graph-based semi-supervised classification, demonstrating superior performance on benchmark datasets like Cora, Citeseer, and Pubmed. It includes command-line arguments for easy configuration of training parameters and model options.

ImageCaptioning.pytorch is a comprehensive open-source codebase designed for advanced image captioning research. It offers robust support for self-critical training, a technique crucial for optimizing caption generation. Researchers can leverage bottom-up features for more detailed image understanding and utilize multi-GPU training for efficient model development, including DistributedDataParallel with pytorch-lightning. The codebase also supports Transformer captioning models, providing a flexible framework for experimenting with state-of-the-art architectures. It includes functionalities for evaluating models on various datasets like COCO and Flickr30k, generating captions for raw images, and performing beam search for improved decoding. With detailed instructions for installation, data preparation, and training, it serves as a valuable resource for academics and developers in the field of computer vision and natural language processing.

LLMRec is a novel framework implemented in PyTorch, designed to significantly improve recommendation systems through the application of three distinct LLM-based graph augmentation strategies. These strategies include reinforcing user-item interactive edges, enhancing item node attributes, and conducting user node profiling, all from a natural language perspective. The tool leverages content within online platforms like Netflix and MovieLens to augment interaction graphs. It provides code, original data, and augmented data, making it a valuable resource for researchers and data scientists working on recommendation systems. LLMRec also offers multi-modal datasets, including textual and visual data, and supports LLM-augmented textual data and embeddings for comprehensive research.

OpenFace is a state-of-the-art, open-source toolkit designed for comprehensive facial behavior analysis. It enables real-time facial landmark detection, accurate head pose estimation, robust facial action unit recognition, and precise eye-gaze estimation. Developed by Tadas Baltrušaitis in collaboration with CMU MultiComp Lab, OpenFace is intended for computer vision and machine learning researchers, as well as the affective computing community. The tool stands out for its ability to run efficiently from a simple webcam without requiring specialized hardware, making advanced facial analysis accessible. It provides source code for both running and training models, ensuring flexibility and extensibility for research and application development.