Research & Education

The Army Research Laboratory (ARL) EEGModels project offers a robust collection of Convolutional Neural Network (CNN) models specifically designed for EEG signal processing and classification. Built with Keras and Tensorflow, this open-source tool aims to support reproducible research by providing well-validated models like EEGNet (including its SSVEP variant), DeepConvNet, and ShallowConvNet. Researchers can easily import and configure these models for their data, compile them with appropriate loss functions and optimizers, and then fit and predict on new test data. The project also includes guidance on feature explainability using tools like DeepExplain, making it a comprehensive resource for deep learning applications in electroencephalography.

Spektral is an open-source Python library designed for graph deep learning, leveraging the Keras API and TensorFlow 2. It offers a straightforward yet adaptable framework for developing Graph Neural Networks (GNNs). The library supports a wide array of popular convolutional layers, such as GCN, Chebyshev, GraphSAGE, ARMA, ECC, GAT, APPNP, GIN, and Diffusional Convolutions, alongside various pooling layers like MinCut, DiffPool, Top-K, SAG, Global, Global gated attention, and SortPool. Spektral also provides extensive utilities for representing, manipulating, and transforming graphs, making it suitable for tasks like classifying social network users, predicting molecular properties, generating graphs with GANs, and clustering nodes. The 1.0 release introduced standardized Graph and Dataset containers, a new Loader class for batching, a transforms module, and GeneralConv/GeneralGNN classes for simplified model building.

vnet.pytorch offers a PyTorch implementation of V-Net, a fully convolutional neural network specifically designed for volumetric medical image segmentation. This open-source project enables researchers and developers to apply V-Net to medical imaging tasks, with a particular focus on segmenting lungs from the LUNA16 dataset. The implementation includes features such as batch normalization and dropout, and provides the option to use NLLoss in addition to the Dice Coefficient for loss calculation. It also includes scripts for generating compute graphs and is derived from established PyTorch repositories, ensuring a robust foundation for medical image analysis.

The Embedded Vision Systems Group, part of the Computer Vision Laboratory at AGH University, is dedicated to advanced research in computer vision. Their primary areas of focus include real-time image and video processing, developing robust embedded vision systems, and implementing control mechanisms for autonomous vehicles. Additionally, the group actively explores applications of deep neural networks and specializes in medical image processing, contributing to both theoretical advancements and practical solutions within these fields. Their research aims to push the boundaries of visual perception and intelligent systems.

Intelimel offers AI-driven physics-based digital twin solutions for engineering and manufacturing. Unlike many digital twin technology companies that provide platforms or toolkits, Intelimel delivers outcome-focused solutions as deployable engineering applications. It integrates AI across multiple levels, including machine learning models built on historical process data for rapid predictions, and physics-guided models combining simulation data (CFD, DEM, FEA) with AI to capture process behavior and spatial effects. The platform also uses agentic AI for intuitive application interaction and deeper insights through natural queries. Intelimel's approach brings together governing physics, process historian data, and visual observations into a single, structured, and explainable model to improve product quality, increase throughput, and reduce waste.

DeepPurpose is a comprehensive deep learning toolkit designed for molecular modeling and prediction in life science research. It supports various applications including Drug-Target Interaction (DTI) prediction, Compound Property Prediction, Protein-Protein Interaction (PPI) prediction, Drug-Drug Interaction (DDI) prediction, and Protein Function prediction. Built on PyTorch, DeepPurpose offers over 15 powerful encodings for drugs and proteins, ranging from cheminformatics fingerprints to graph neural networks, providing over 50 combined models. The library is designed for realistic and user-friendly applications, supporting tasks like drug repurposing, virtual screening, QSAR, and side effect prediction. It includes features like automatic task identification (regression/binary), support for cold target/drug settings, extensive dataset loading scripts, pretrained checkpoints, and detailed training metrics with early stopping. DeepPurpose is actively maintained and encourages user feedback.

The Meta Agents Research Environments Demo is a web application designed to provide a platform for exploring and interacting with research environments developed for Meta agents. Users can easily access the site and select an environment to view its details or initiate a simulation. This tool is particularly valuable for researchers and machine learning engineers who are interested in understanding and experimenting with AI agent behaviors within various simulated settings. It offers a straightforward way to engage with advanced AI research without requiring complex setup or specialized input, making it accessible for both detailed analysis and general exploration of agent capabilities.

Flora Incognita is an AI-powered mobile application designed for interactive plant species identification. Users can identify over 30,000 plant species, including wild herbs, trees, grasses, cacti, palms, ferns, and some cultivated plants, by simply taking photos with their smartphone camera. The app also identifies about 500 common moss species and, since 2025, approximately 3,000 lichen and fungus species. It provides extensive plant fact sheets with information on characteristics, protection status, and distribution. Flora Incognita is free of charge and advertising, and it functions offline, making it an ideal tool for educational purposes in schools, universities, and nature conservation initiatives. It supports citizen science by allowing users to save observations, contributing valuable data for biodiversity monitoring and research.

PINNs (Physics Informed Neural Networks) is an open-source deep learning framework designed to solve supervised learning tasks while adhering to physical laws described by nonlinear partial differential equations. It offers capabilities for both data-driven solution and data-driven discovery of PDEs. The tool supports continuous time and discrete time models, forming a class of data-efficient universal function approximators that embed underlying physical laws as prior information. PINNs can infer solutions to PDEs, create physics-informed surrogate models, and facilitate the discovery of partial differential equations from data. While the original repository is no longer actively maintained, the underlying concepts are widely implemented in PyTorch, JAX, and TensorFlow v2.

PointNet++ is a deep learning framework designed for hierarchical feature learning on point sets, building upon and extending the original PointNet architecture. It addresses the challenge of non-uniform densities in natural point clouds by proposing special layers that intelligently aggregate information from different scales. The framework learns hierarchical features with increasing scales of contexts, similar to convolutional neural networks. This repository provides code and data for PointNet++ classification and segmentation networks, along with utility scripts for training, testing, data processing, and visualization. It is implemented in TensorFlow and supports multi-GPU training, making it suitable for researchers and engineers working with 3D point cloud data.

SentinelOne is an AI-powered tool designed for climate risk assessment and monitoring, available as a Hugging Face Space. It leverages AI agents to analyze location-specific data and generate comprehensive risk assessment reports. Users provide their area of interest, and the application processes this information to identify and evaluate potential climate-related risks. This tool is particularly useful for researchers, environmental agencies, and anyone needing to understand the climate vulnerabilities of a specific geographical area, offering a streamlined approach to complex environmental data analysis.

Simple Vectorization is a tool hosted on Hugging Face Spaces, designed for quickly generating vector embeddings. It serves as a valuable resource for educational purposes, allowing users to experiment with fundamental AI concepts related to vectorization. The tool is freely accessible, making it an ideal platform for students, researchers, and enthusiasts to explore and understand how data can be transformed into numerical vectors for machine learning applications. While the live website currently shows a runtime error, its intended function is to provide a straightforward way to engage with vectorization processes.

TerraEye leverages AI and satellite data processing to enhance mineral exploration and mine monitoring. The tool offers spectral targeting capabilities, allowing for more precise identification of mineral deposits. It also supports environmental surveillance, impact assessment, and rehabilitation monitoring for mining operations. By utilizing geo-spatial analytics and remote sensing, TerraEye aims to reduce costs and improve the efficiency of mineral exploration, providing valuable insights for various stages of mining, from initial discovery to environmental management.

Video Model Studio offers an all-in-one solution for AI video training, providing a Gradio-based interface for comprehensive model management. Users can upload and process videos, train models, and manage storage directly within the application. This tool is designed to streamline the workflow for developers and researchers working with AI video, facilitating both video analysis and generation research. It aims to simplify the complex process of fine-tuning video models through an accessible interface.

XTTS Voice Clone on CPU is a Hugging Face Space that enables users to generate realistic synthesized speech by inputting text and a short audio clip. This tool is designed for voice cloning, allowing users to create custom voices in their chosen language. It supports both uploading reference audio and using a microphone for input. While the tool itself is hosted on Hugging Face Spaces, which offers a free tier for basic CPU usage, more advanced hardware and dedicated inference endpoints are available through Hugging Face's paid plans. This makes it accessible for experimentation while also providing options for scaling up.

WebLLM Structured Generation Playground is an innovative AI tool hosted on Hugging Face Spaces, designed for experimenting with structured data generation. Users can provide a text prompt, select an LLM model, and define a JSON schema or custom EBNF grammar. The tool then runs the chosen model directly within the user's browser, ensuring that the generated output strictly adheres to the specified structure. This capability is invaluable for developers, AI researchers, and LLM enthusiasts who need to test and refine AI models for producing consistent, structured outputs. It offers a hands-on environment to understand and control the output format of large language models, making it a powerful resource for advanced AI development and research.

Voice Conversion Yourtts is an AI tool designed for voice conversion, leveraging the Yourtts technology. It provides a platform for researchers and developers to experiment with and implement voice cloning techniques. The tool is particularly useful for those looking to create custom voices or develop voice-based applications. While the specific features are not detailed, its focus on voice conversion and cloning suggests capabilities for transforming audio inputs into different voices. The platform is hosted on Hugging Face Spaces, indicating an environment for machine learning applications. However, at the time of scraping, the application was experiencing a runtime error due to memory limits, suggesting potential resource intensity.

🗣️ASR Clone Voice AI Gradio🔊 is an AI-powered voice cloning tool available on Hugging Face Spaces. It leverages Automatic Speech Recognition (ASR) technology to enable users to clone voices. While the tool's specific features beyond voice cloning are not detailed, its presence on a platform like Hugging Face suggests it is likely accessible for experimentation and development within the AI community. The current status indicates a build error, meaning it is not functional at this time.

SmallVill offers a captivating virtual world where 25 AI agents, each inspired by historical figures like Socrates and Cleopatra, engage in dynamic conversations and actions within a modern-day village setting. Users can observe the unfolding lives of these AI characters, providing a unique simulation experience. Beyond the interactive virtual world, SmallVill also features exclusive NFTs available on OpenSea, blending AI simulation with digital collectibles. This platform is ideal for those interested in observing complex AI behaviors and interactions in a simulated environment.

Blocks is an open-source framework built on top of Theano, designed to simplify the construction and training of neural networks. It offers several key features including the ability to create 'bricks' for parametrized Theano operations, pattern matching for selecting variables and bricks within complex models, and algorithms for model optimization. The framework also supports saving and resuming training sessions, monitoring and analyzing training progress across different datasets, and applying graph transformations like dropout. Blocks is complemented by Fuel, a data processing engine, and has additional components available through Blocks-extras, making it a comprehensive solution for deep learning development.

Solarad AI offers an advanced AI-powered platform for solar energy forecasting and weather intelligence, designed to maximize solar energy assets and minimize grid penalties. The platform provides 14-day solar power forecasts with sub-hourly resolution, utilizing ensemble ML and NWP models, including plane-of-array (POA) for PV trackers. Users can leverage hyperlocal monitoring, predictive insights, and automated alerts to make informed operational decisions. Solarad AI also offers robust Weather APIs for Solar, enabling real-time and historical solar weather data integration. With a focus on leading accuracy, the platform provides over 50 data layers and actionable insights to safeguard plant operations and deliver timely alerts for weather disruptions.

BluMotiv specializes in smart, scalable, and sustainable electric mobility solutions, leveraging AI to enhance electric vehicle performance. The platform provides advanced EV powertrain solutions and electrofit services, enabling the conversion of conventional vehicles into smart EVs. By utilizing AI-powered predictive controls, BluMotiv optimizes critical aspects such as vehicle performance, extended range, and improved battery life. The company also employs virtual twin technology and AI-based insights, which are crucial for the development and refinement of their EV solutions, ensuring efficiency and sustainability in the evolving electric vehicle market.

ARQUIMEA Research Center is a corporate research center dedicated to developing ideas and projects with high technological value. It focuses on disruptive technologies including quantum computing, biotechnology, photonics, robotics, artificial intelligence, and blockchain. The center is part of ARQUIMEA, a tech company that invests a significant portion of its annual profits into R&D to create new products and services. It employs researchers and experts with extensive experience from various global research centers, aiming to solve societal problems through science and technology and contribute to the development and progress of society across high-demanding sectors.

BioGPT is an open-source generative pre-trained transformer specifically designed for biomedical text generation and mining. Developed by Microsoft, it offers pre-trained models and fine-tuned checkpoints for a range of biomedical tasks. Researchers can leverage BioGPT for applications such as question answering on PubMedQA, relation extraction on datasets like BC5CDR and DDI, and document classification. The tool is implemented in PyTorch and integrates with the Hugging Face transformers library, making it accessible for use in various research workflows. It supports both general text generation and specialized tasks within the biomedical domain, providing a powerful AI model for scientific text analysis.