Whitepaper 1


Synthetic Human Data For Research

Highly Diversified Dataset

With more than 5000 individually photo scanned and optimized human 3D models there is no limit for complex machine learning applications. Whether it's 3D pose estimation, computer vision training, or annotation. Due to the amount of the datasets and the diversity of ethnicities, ages, clothing styles and poses there is much data to work with.

Detailed Tagging and Metadata

The human models included in our dataset have been scanned and created with categorization and metadata in mind. Our Human Datasets include all relevant object information to complement our ground truth 3D data. With detailed additional information about age, body height, clothing style, pose, or activity you get all you need to train your CV models.

Ground Truth Photoscan Data

All human 3D models are photo scanned copies of real-world human actors. Capturing all visual details to create 1:1 synthetic human data copies is key get ground truth training sets for any application. 8K texture resolution with pore-level detail and different alpha masks for every surface of the model deliver rich and extensive visual information.

Unified and Optimized Data Structure

With a focus on research, ease of use of datasets is always key to streamline ML pipelines. All textures, 3D data, and related files come with a standardized and ready-to-use naming scheme, folder structure, as well as image and mesh resolution. Ensuring that all of our models are comparable and easy to integrate into ML learning models.

What are Human Datasets?

Human datasets are used to train machine learning models or computer vision research algorithms. Ground truth accuracy and high visual detail of the human 3D scan data are key to create reliable training results.

Our Human dataset contains more than 5000 individually photo scanned human 3D models in different poses and clothing styles. Optimized and ready to use for ML, CV, and AI applications.