Research & Education

Federated Learning with Substra is an open-source platform designed for federated learning research and development. It facilitates secure data analysis and collaborative model training, allowing multiple parties to train a common model without sharing their raw data. The platform leverages technologies like Gradio for its interface and is licensed under GPL-3.0, promoting community contributions and transparency. While the current live website indicates a runtime error, the underlying purpose is to provide a robust environment for advancing federated learning techniques, which is crucial for privacy-preserving AI development.

Ferret Demo is an AI model demonstration tool hosted on Hugging Face Spaces, developed by Jade Choghari. It enables users to upload an image and provide a text prompt to receive a detailed description of the image's contents. A key feature is the ability to draw a bounding box on the image, allowing users to focus the AI's attention on specific areas for more precise analysis. This tool is designed for exploring and testing AI capabilities related to image understanding and description. While the demo currently experiences runtime errors due to workload eviction and storage limits, its core functionality aims to provide a platform for AI enthusiasts, researchers, and developers to experiment with image-to-text models.

Gpu Tflop Finder is a specialized tool hosted on Hugging Face Spaces, designed to provide quick access to GPU TFLOPS data. Users can easily view and filter this information by categories such as consumer, workstation, or datacenter GPUs. The interface includes toggle checkboxes, enabling users to hide or show specific GPU types, streamlining the search for relevant performance metrics. This tool is particularly useful for AI developers and researchers who need to evaluate GPU performance for hardware selection and optimizing AI models, offering a straightforward way to compare different GPUs based on their TFLOPS.

Gradio_opencv is a specialized tool designed to bridge the gap between OpenCV's powerful computer vision capabilities and Gradio's user-friendly interface for machine learning applications. It enables developers and researchers to easily create interactive web demos for image processing and computer vision tasks. The tool facilitates the integration of complex OpenCV functions into Gradio applications, making it simpler to showcase and test computer vision models. This is particularly useful for those working on real-time video analysis or developing prototypes that require visual interaction. While the current live website indicates a runtime error, the core purpose of Gradio_opencv is to streamline the development and deployment of computer vision applications within the Gradio ecosystem.

AI-Optimizer is a comprehensive deep reinforcement learning toolkit developed by TJU-DRL-LAB. It offers a wide array of algorithm libraries, spanning from model-free to model-based RL, and supports both single-agent and multi-agent reinforcement learning. The toolkit also includes a flexible and easy-to-use distributed training framework designed for efficient policy training. Key areas of focus include Multiagent Reinforcement Learning (MARL), Offline Reinforcement Learning (OffRL), Self-supervised Reinforcement Learning (SSRL), and Transfer and Multi-task Reinforcement Learning. It aims to address challenges like the curse of dimensionality, non-stationarity, and sample inefficiency in RL, providing solutions for researchers and practitioners alike.

Prithvi 100M Sen1floods11 is a demonstration tool developed by IBM-NASA Geospatial, designed for analyzing flood data using artificial intelligence. Users can upload Sentinel-2 image files, which must contain all 12 spectral bands and be scaled by 10,000. The application then processes these images to return an original RGB picture alongside a black-and-white mask. In this mask, white areas indicate water, while black areas represent land. This tool is particularly useful for exploring geospatial data and testing AI models related to flood detection and environmental monitoring. It operates as a web application, making it accessible for various research and analytical purposes.

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.

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.

Unet is an open-source implementation of the U-Net deep learning framework, built with Keras. It is specifically designed for image segmentation tasks, drawing inspiration from convolutional networks used in biomedical image segmentation. The tool provides a robust foundation for developers and data scientists to build and train their own image segmentation models. It includes pre-processed data from the ISBI challenge, data augmentation capabilities using Keras's ImageDataGenerator, and a model implemented with Keras functional API. The network outputs a 512x512 mask with pixel values in the [0, 1] range, using a sigmoid activation function. The model is trained with binary crossentropy as the loss function and achieves high accuracy after a few epochs.

WAR GAME AI Simulation is an AI-powered tool designed for simulating military battles and strategic scenarios. Users can customize their simulations by selecting various terrains, nations, and unit types. The platform allows for the creation of specific missions, providing detailed analysis of combat power and offering control over battle speed. It enables users to deploy forces, initiate engagements, and review comprehensive battle reports. This tool is ideal for educational purposes, strategic planning, and understanding military tactics through interactive simulations.

The python-machine-learning-book repository serves as the official code and information resource for the first edition of the "Python Machine Learning" book. It provides over 400 pages of useful material, covering everything from machine learning theory to practical code implementations using NumPy, scikit-learn, and Theano. The resource aims to explain underlying concepts, best practices, and caveats, rather than just demonstrating how scikit-learn works. It includes code notebooks for each chapter, excerpts from the foreword and preface, setup instructions for Python and Jupyter Notebook, and additional math and NumPy resources. The repository also features bonus notebooks, related content, and slides for teaching, making it a comprehensive learning companion.

Aigenpulse.com is a domain name currently listed for sale on HugeDomains.com. The domain can be purchased outright for $4,995 or financed through a payment plan of $208.13 per month for 24 months with 0% interest. HugeDomains offers a 30-day money-back guarantee and promises quick delivery of the domain, typically within one to two hours of purchase during business hours. The purchase includes only the domain name, with no additional services like hosting or web design. Buyers can transfer the domain to any registrar after purchase, though payment plan domains are not transferable until fully paid. WhoIs Privacy Protection is available through NameBright.com, the registrar where the domain is pushed after purchase.

This repository, deep-learning-localization-mapping, serves as a comprehensive collection of deep learning-based localization and mapping approaches. It includes models for various tasks such as odometry estimation (visual, visual-inertial, inertial, LIDAR), geometric and semantic mapping, and global localization. The repository also features survey papers on deep learning for visual localization and mapping, and deep learning for inertial positioning, providing a valuable resource for understanding the state-of-the-art in spatial machine intelligence. Researchers and engineers in robotics, computer vision, and related fields will find this collection useful for exploring and implementing advanced localization and mapping techniques.

AnalysisMode offers AI solutions to unlock the full potential of bioprocessing, primarily through its SimCell platform. SimCell acts as a virtual cell culture lab, allowing users to design and run virtual experiments to predict growth patterns, productivity, and process variability. This accelerates R&D cycles, reduces the need for wet-lab work, and provides data-driven insights for optimizing critical process parameters. Beyond SimCell, AnalysisMode provides tailored AI services covering data assessment, AI-driven Design of Experiments (DoE), digital twin creation for scale-up, and continuous AI feedback loops for real-time monitoring. The platform aims to minimize variability, maximize reliability, and significantly reduce development timelines and costs in bioprocessing.

Nabla Bio is at the forefront of drug development, leveraging generative drug design combined with extensive, patient-relevant testing to transform the discovery process into a precise engineering discipline. The platform integrates data, AI, and both dry and wet-lab systems into a unified engine, enabling the design of medicines with optimized properties such as binding, developability, cellular function, and in-vivo performance. This approach aims to create new medicines that are safer, more precise, and can unlock novel therapeutic targets, ultimately expanding our understanding and treatment of diseases. Nabla Bio partners with leading pharmaceutical companies to maximize the impact of its platform and bring these advanced drug candidates to patients faster.

VRP-RL is an open-source project that leverages reinforcement learning to tackle complex combinatorial optimization problems such as the Vehicle Routing Problem (VRP) and the Traveling Salesman Problem (TSP). Developed by OptMLGroup, this tool provides a robust framework for researchers and developers to implement, train, and evaluate reinforcement learning models for route optimization. It is built using TensorFlow and includes all necessary dependencies like NumPy and tqdm. Users can easily run the code for both training and inference, with options to specify GPU usage, model directories, and inference types (batch or single mode). The project also logs all results, making it suitable for experimental research and performance analysis in the field of operational research and artificial intelligence.

Segmentation Of Teeth In Panoramic X Ray Image Using U Net is an AI-powered tool designed for the automatic segmentation and highlighting of teeth within panoramic X-ray images. Utilizing a U-Net architecture, the application processes uploaded X-ray images to accurately identify and delineate individual teeth. The segmented teeth are then overlaid in red on the original image, providing a clear visual representation. This capability is particularly beneficial for dental professionals, researchers, and students, as it streamlines the analysis of X-ray images, assists in diagnostic processes, and supports dental research by automating a crucial aspect of image interpretation. The tool is accessible via a web interface, allowing users to easily upload images and receive processed results.

SegFormer (ADE20k) in TensorFlow is an AI tool specifically designed for semantic image segmentation. Built with TensorFlow, it enables detailed image analysis and object recognition, making it suitable for tasks that require precise pixel-level classification. This tool is particularly useful for researchers and developers working in computer vision who need to accurately identify and delineate different objects or regions within an image. Its implementation within the TensorFlow framework ensures compatibility with a wide range of machine learning workflows and environments, facilitating integration into existing projects.

Sesame CSM is a conversational speech generation tool hosted on Hugging Face Spaces, designed to create realistic dialogue between two distinct speakers. Users can input brief text descriptions and optional audio samples to define each speaker's voice. Following this setup, a dialogue can be typed out with alternating lines for each speaker. The application then processes this input to generate a single, cohesive audio file that voices the entire conversation, making it suitable for various applications requiring multi-speaker audio output. It's an accessible tool for generating conversational speech without complex setups.

deep-rl-tensorflow offers a TensorFlow implementation of several key deep reinforcement learning papers, making advanced algorithms accessible for research and development. This open-source project includes implementations of foundational works such as 'Playing Atari with Deep Reinforcement Learning' and 'Human-Level Control through Deep Reinforcement Learning,' alongside more recent advancements like Double Q-learning and Dueling Network Architectures. It also features in-progress implementations for Prioritized Experience Replay, Deep Exploration via Bootstrapped DQN, Asynchronous Methods for Deep Reinforcement Learning, and Continuous Deep Q-Learning with Model-based Acceleration. The tool provides clear usage instructions for training models with different network configurations and environments, making it a valuable resource for researchers and engineers working on reinforcement learning projects using TensorFlow.

Splatt3R is an AI-powered tool hosted on Hugging Face Spaces that enables zero-shot Gaussian splatting from uncalibrated image pairs. Users can easily upload one or two images, and the application will process them to generate a 3D model in PLY file format. This model can then be viewed directly within the application or downloaded for further rendering and manipulation in other 3D viewers and software. The tool provides an accessible way to experiment with AI for creating three-dimensional representations from standard images, making advanced 3D modeling techniques available to a broader audience without requiring specialized calibration equipment.

SuperGlue Image Matching is an AI tool hosted on Hugging Face Spaces, designed for identifying corresponding features between different images. This capability is crucial for various computer vision tasks such as object recognition and visual localization. While the specific application details are not extensively provided on the live page, its presence on Hugging Face suggests it leverages advanced machine learning models for robust image analysis. The platform itself offers various pricing tiers for compute resources, allowing users to scale their usage based on their needs, from free CPU options to powerful GPU instances for more demanding tasks. This makes it accessible for both individual researchers and larger teams working on complex AI projects.