Data & Analytics

Lightwheel is an infrastructure company specializing in Physical AI, providing a comprehensive data engine for world models and robotic learning. It offers three core layers: the World Layer with its SimReady Library for high-quality, physically accurate assets and scenes; the Behavior Layer featuring EgoSuite for globally scalable egocentric human data; and the Evaluation Layer with RoboFinals, an industrial-grade simulation platform for challenging VLA and world models. The Lightwheel-Platform Enterprise unifies these components into a single stack for building, training, and deploying Physical AI at scale, enabling robots to learn efficiently.

One Panel is an innovative web-based Manga Reader designed to transform the traditional manga reading experience. It allows users to read their favorite manga one panel at a time, eliminating spoilers and building tension. This unique approach reimagines how manga is consumed, encouraging readers to dedicate proper time and attention to each panel. The tool offers a completely new experience where panels can be inserted or removed with a single keystroke, making it ideal for manga lovers who want to savor every moment. It's particularly useful for streamers or content creators who wish to build anticipation and prevent spoilers during live sessions, ensuring the page is built gradually rather than revealed all at once.

Strand AI specializes in generating missing multimodal patient data, including gene expression, proteomics, and spatial transcriptomics, from routinely collected samples. This capability is crucial for life sciences teams engaged in clinical trials and biomarker discovery, where incomplete patient cohorts or unmeasured modalities can hinder research. The platform helps rescue incomplete cohorts by predicting missing data, allowing researchers to avoid discarding valuable subjects. It also enables the prediction of expensive assays like proteomics or transcriptomics from readily available H&E slides and genotypes. Strand AI is particularly useful for unlocking rare disease cohorts by filling in missing modalities, ensuring sufficient data for model training, and for discovering unmeasured biomarkers across entire cohorts.

YEDDA (formerly SUTDAnnotator) is a lightweight and collaborative text span annotation tool designed for annotating chunks, entities, and events in text, symbols, and even emojis. It supports a wide range of languages, including English and Chinese. The tool offers highly efficient annotation through shortcut keys, allowing users to select text spans and apply labels with a single key press. It also features a command annotation model for batch processing multiple entities and intelligent recommendation to suggest entities based on annotated text. YEDDA provides both an annotator interface for efficient labeling and an administrator interface for quality analysis, including multi-annotator analysis and pairwise comparisons. It is compatible with Windows, Linux, and MacOS, and allows for seamless switching and customization of shortcut maps.

MacV AI leverages existing camera infrastructure to provide real-time vision AI for industrial safety and retail analytics. The platform connects to RTSP camera streams, detecting violations like PPE non-compliance or zone breaches with approximately 95% accuracy. It instantly alerts supervisors via SMS, email, or API within seconds of detection. Beyond immediate alerts, MacV AI analyzes patterns over time to identify systemic risk factors, offering insights that traditional methods miss. It's designed for enterprise, compatible with any IP camera, NVR, or DVR system, and processes data locally with edge AI for privacy and security, ensuring compliance with regulations like GDPR.

Photofeeler offers a unique service for obtaining unbiased feedback on your photos, helping you choose the best images for various online profiles. Users upload photos and receive scores on key traits from real people, along with qualitative feedback to improve their online image. The platform supports testing for business, social, and dating contexts, measuring traits like attractiveness, trustworthiness, competence, and likability. Users can earn free feedback by voting on others' photos or purchase credits for faster results. Photofeeler emphasizes privacy and control, ensuring photos are only visible during active tests and are never indexed by search engines or used without consent. Its sophisticated AI and fraud detection systems ensure the accuracy and integrity of test results.

Kaleidoo is an AI solution platform that is currently undergoing a significant transformation. The company is actively growing, evolving, and constructing a brand new website to showcase its updated offerings. Users are encouraged to stay tuned for the upcoming launch, which promises exciting new developments and features. While specific details about the platform's capabilities are not yet available, the previous description indicated its focus on integrating data and workflows into a modular AI platform, providing actionable insights for industries like healthcare, retail, and finance. It previously supported data labeling, machine learning, video analytics, and NLP.

SmartOne.ai offers specialized real-world AI data services, including annotation, evaluation, and domain expertise for systems that operate outside controlled lab environments. The platform leverages advanced automation combined with expert human insight (Human-in-the-Loop) to precisely categorize, classify, and annotate various data types such as text, images, audio, and video. Key services include Physical AI Data Annotation for embodied intelligence, Data Labeling for AI across industries, and Synthetic Data generation that mirrors real-world conditions without sensitive data challenges. SmartOne.ai also provides Smart Edge Specialists to handle complex technical challenges in AI, IoT, and next-generation technologies, empowering teams to focus on strategy.

U-Mamba is an open-source tool designed to enhance long-range dependency for biomedical image segmentation, built upon the popular nnU-Net framework. It provides a comprehensive repository for researchers and developers focused on deep learning applications within the biomedical field. The tool supports Pytorch 2.0.1 and Mamba, facilitating the implementation of advanced image analysis techniques. Users can train both 2D and 3D models, including U-Mamba_Bot and U-Mamba_Enc, on their own datasets by following provided guidelines for data preparation and preprocessing. It also offers functionalities for inference on testing cases and allows for flexible path settings for existing nnU-Net setups.

Datamundi specializes in providing high-quality, ethically sourced human data for AI, encompassing speech, text, image, and video. Their services range from data collection and labeling to expert data management, ensuring bias-aware training data that scales with AI needs. Beyond AI data, Datamundi offers comprehensive language services for culturally appropriate messaging across global markets, and managed services to handle complexities of international communication. They emphasize a human-driven, AI-smart approach, building strong client partnerships to transform multilingual data challenges into competitive advantages. The company, operating under the legal name Summa Linguae Technologies, focuses on agility and precision to deliver scalable data solutions.

PixelAnnotationTool is an open-source software designed for manual and quick image annotation, particularly useful for computer vision tasks. It employs a pseudo-manual approach leveraging OpenCV's watershed algorithm, where users provide initial markers with brushes, and the algorithm refines the segmentation. If corrections are needed, users can easily refine markers on erroneous areas. The tool supports annotating images within directories and is available for Linux, Mac, and Windows. It requires dependencies like Qt, CMake, and OpenCV for building, and binaries are available for download. The project is hosted on GitHub, encouraging community contributions and offering a donation option for maintenance and updates.

Reliabl is a collaborative data annotation platform designed to transform human input and user feedback into reliable human-in-the-loop AI training data. It enables users and experts to efficiently create high-quality, contextual, and reliable AI training data together, helping to prevent AI data bias and ensure ethical, trustworthy AI performance. The platform allows for custom, context-specific labels and embraces data complexity by bringing together diverse perspectives to navigate nuances and interpretations through custom taxonomies. Reliabl streamlines the entire annotation workflow, from data collection and flexible annotation schemas to preparation for model training, all within one integrated, research-backed system. It also prioritizes user data privacy through transparency, consent, and ethical handling of information, including robust security measures like data anonymization and HIPAA compliance.

Dori AI is a visual intelligence platform designed to empower businesses with real-time visibility and operational insights, particularly within manufacturing and logistics. The platform leverages computer vision for critical tasks such as quality inspection and productivity monitoring. By providing real-time analytics, Dori AI helps companies automate inspections, improve process adherence, and ultimately optimize their operations. It aims to enhance efficiency and decision-making through advanced visual intelligence capabilities, making it a valuable asset for businesses looking to streamline their industrial processes and gain deeper insights into their operational performance.

AIMMO is an AI data company specializing in optimizing small-scale data for high-performance AI. The platform offers an end-to-end service covering data collection, curation, labeling, and synthetic data generation. Key offerings include AIMMO Core™ Data Managed Service, which provides customized solutions developed by industry specialists, and AIMMO Core™ AI Solution Service, which delivers tailored AI models, multi-sensor data collection systems, Edge AI, and Data Platforms. AIMMO also has an AIaaS (AI as a Service) offering in preparation, which will provide subscription-based access to its AI capabilities via web applications, APIs, and SDKs. The company focuses on efficient data optimization through consistent labeling of small datasets, edge case data augmentation, and traceable processes for data and model reuse, aiming to reduce manual effort and production costs.

Bossa Nova was a pioneer in automated on-shelf inventory management, utilizing robots and AI to monitor retail environments. Their technology was successfully deployed and scaled in 600 Walmart Supercenters, where robots scanned aisles multiple times daily to identify and report operational issues such as out-of-stock products. This extensive experience provided significant insights into retail operations and AI deployment. The company now dedicates its website to sharing these learnings, offering resources and advice for startups and new entrepreneurs, covering topics from strategy frameworks and fundraising to specific challenges in robotics and retail.

DataClap offers end-to-end services for training data, from data collection to annotation, for AI, ML, and computer vision companies. They provide fully managed services with a scalable workforce and a two-layer quality control process to ensure high accuracy. DataClap supports a wide range of annotation types, including classification, bounding box, polygon, cuboid, point cloud, and key points for text, image, video, audio, and LiDAR data. Their process involves consultation, a free pilot project, feedback, customized workflow planning, annotation, and a final QC and delivery stage. DataClap emphasizes data security, being GDPR compliant and ISO 27001 certified, and offers cost-efficient solutions compared to contract workers or large annotation platforms.

Humans in the Loop is a social enterprise offering comprehensive data annotation services for AI, focusing on ethical practices and high data accuracy. They provide a wide range of annotation types including bounding box, polygon, keypoint, semantic segmentation, medical, video, and 3D annotation. The service supports various industries such as geospatial, automotive, medical, industrial, agricultural, and retail AI. A key differentiator is their social mission to provide digital work and training to conflict-affected communities, ensuring fair pay and career development. They also offer solutions for dataset collection, human-in-the-loop for active learning, real-time edge case handling, and reinforcement learning with human feedback, ensuring secure and GDPR-compliant data handling.

CLIP Caption Reward is a specialized tool designed for evaluating the quality of image captions, leveraging the power of CLIP (Contrastive Language-Image Pre-training) to provide a reward signal. This makes it particularly useful for researchers and developers working on image captioning models, allowing them to assess and refine their generated captions. Built with Gradio, it offers an interface for interaction, though the current live version indicates runtime errors preventing full functionality. The tool aims to support AI research and development in the domain of image understanding and generation, specifically focusing on the accuracy and relevance of textual descriptions for images.

Depth Anywhere App is an AI-powered tool available as a Hugging Face Space that allows users to easily generate depth maps from uploaded images. By simply providing a photograph, the application processes it to produce an output image that visually represents the distances of objects within the scene. This functionality is particularly useful for tasks requiring 3D understanding from 2D inputs, such as computer vision research, 3D reconstruction, or artistic applications. The tool is built with Gradio, making it accessible through a web interface, and is licensed under Apache-2.0, indicating its open-source nature suitable for various development and educational purposes.

go-face is an open-source Go library designed for face recognition, integrating with the powerful dlib machine learning toolkit. It enables developers to implement robust face detection and recognition functionalities directly within their Go applications. The library requires specific dlib and libjpeg development packages for compilation across various operating systems like Ubuntu, Debian, macOS, and Windows (via MSYS2). It supports the use of pre-trained models for face landmarks, human face detection, and face recognition, which can be downloaded from a companion repository. The tool offers detailed GoDoc documentation and provides clear examples for initialization, face recognition from images, and classification of recognized faces.

Ximilar offers a robust AI platform designed for businesses to enhance their image processing capabilities through advanced image recognition and visual search APIs. The platform automates tasks such as image tagging, description generation, sorting, and searching, significantly reducing manual effort and costs. It supports various applications, including product recommendations in e-commerce, content curation, and identification of collectibles like stamps, coins, and comic books. Ximilar's solutions are built to handle large datasets, processing millions of images efficiently while prioritizing data security and compliance with regulations like GDPR. Developers can access its capabilities via REST API, with support for custom model training and continuous optimization.

img2vec is an open-source Python library that leverages pre-trained PyTorch models to extract vector embeddings from images. This tool is highly valuable for various image analysis tasks, including ranking for recommender systems, clustering images into different categories, and classification tasks. It supports a wide range of models such as Resnet-18, Resnet-50, Alexnet, Vgg, Densenet, and EfficientNet, each returning a vector of a specific length. Users can easily integrate img2vec into their projects, specifying whether to run on GPU, choose a particular model, and even select a specific layer for output extraction. This flexibility makes it a powerful solution for developers and data scientists working with image feature extraction.

LiteFlowNet is an open-source, lightweight convolutional neural network specifically engineered for optical flow estimation. Released as a spotlight paper at CVPR 2018, it offers a fast and accurate solution for analyzing motion between video frames. The tool incorporates several specialized modules, including pyramidal features, cascaded flow inference with cost volume and sub-pixel refinement, a feature warping (f-warp) layer, and feature-driven local convolution for flow regularization. LiteFlowNet is known for outperforming other networks like PWC-Net while maintaining a smaller model size, making it suitable for resource-constrained environments. It has evolved into LiteFlowNet2 and LiteFlowNet3, offering further improvements in accuracy and runtime.

HeritCoin is an AI-powered coin identification application designed for numismatists, coin collectors, enthusiasts, and amateurs. The app boasts a 99.9% recognition accuracy, allowing users to quickly identify coins and banknotes by simply taking or uploading a picture. Beyond automated identification, HeritCoin provides access to expert appraisal services for more precise evaluations of currency condition and value. Users can also create 3D collections of their coins and maintain a wishlist. The app supports multiple languages and is available on both the App Store and Google Play, making it a comprehensive tool for managing and learning about coin collections.