World's most popular 3D human datasets
To create ground truth synthetic data for computer vision research and simulation
To create ground truth synthetic data for computer vision research and simulation
Xu Chen; Tianjian Jiang; Jie Song; Jinlong Yang; Michael J. Black; Andreas Geiger; Otmar Hilliges
ETH Zürich; University of Tübingen; Max Planck Institute for Intelligent Systems, Tübingen
Thiemo Alldieck; Mihai Zanfir; Cristian Sminchisescu
Google Research
Salehe Erfanian Ebadi; You-Cyuan Jhang; Alex Zook; Saurav Dhakad; Adam Crespi; Pete Parisi; Steven Borkman; Jonathan Hogins; Sujoy Ganguly
Unity Technologies
Mohamed Hassan; Partha Ghosh; Joachim Tesch; Dimitrios Tzionas; Michael J. Black
Max Planck Institute for Intelligent Systems, Tubingen, Germany
Yasamin Jafarian; Hyun Soo Park
University of Minnesota
Priyanka Patel; Chun-Hao P. Huang; Joachim Tesch; David T. Hoffmann; Shashank Tripathi; Michael J. Black
Max Planck Institute for Intelligent Systems, Tubingen, Germany; University of Freiburg; Bosch Center for Artificial Intelligence
Bindita Chaudhuri; Nikolaos Sarafianos; Linda Shapiro; Tony Tung
University of Washington; Facebook Reality Labs Research, Sausalito
Zeng Huang; Yuanlu Xu; Christoph Lassner; Hao Li; Tony Tung
Facebook Reality Labs, Sausalito, USA; University of Southern California, USA
Shunsuke Saito; Tomas Simon; Jason Saragih; Hanbyul Joo
University of Southern California; Facebook Reality Labs; Facebook AI Research
Tyler Zhu; Per Karlsson; Christoph Bregler
Google Research
Boyi Jiang; Juyong Zhang; Yang Hong; Jinhao Luo; Ligang Liu; Hujun Bao
University of Science and Technology of China; State Key Lab of CAD&CG, Zhejiang University
Thiemo Alldieck; Marcus Magnor; Bharat Lal Bhatnagar; Christian Theobalt; Gerard Pons-Moll
Computer Graphics Lab, TU Braunschweig, Germany; Max Planck Institute for Informatics, Saarland Informatics Campus, Germany
Shunsuke Saito; Zeng Huang; Ryota Natsume; Shigeo Morishima; Angjoo Kanazawa; Hao Li
University of Southern California; USC Institute for Creative Technologies; Waseda University; University of California Berkeley; Pinscreen
Dataset 01
Posed People are lifelike static scans, a 1:1 detailed capture from our scanner including finest details. These 3D models correspond to reality to a very high degree which elevates accuracy and authenticity.
Size: | 3350 Single Models 170 Group Models |
Individuals / Entities: | 203 scanned individual people |
Clothing Variations: | Avg. ~ 16.5 variations per individual |
Polycount: | 100k & 30k resolution (two versions) triangulated |
Texture Size: | 8k & 2k resolution (two versions) |
Textures Included: | Diffuse map Normal map Alpha maps for surface separation |
Formats Available: | OBJ & FBX 3ds Max Cinema 4D Maya SketchUp Rhino Blender |
Dataset 02
Rigged People are rigged scans in A or T pose. So they can be put into every possible pose, or they may even be animated. This is ideal for generating the greatest possible pose variation.
Size: | 793 single models |
Individuals / Entities: | 129 scanned individual people |
Clothing Variations: | avg. ~ 6.15 variations per individual |
Polycount: | ~ 10-15k (retopologized quads) |
Texture Size: | 8k resolution |
Textures Included: | Diffuse map Normal map Gloss Map Alpha maps for surface separation |
Formats Available: | FBX 3ds Max Cinema 4D Maya UE 4 Unity |
HumanDataset is a service by Renderpeople to meet the needs of the computer vision industry. Since 2013 Renderpeople is one of the world leaders in the production, development and distribution of scanned 3D People models as stock footage.
Renderpeople datasets have already been used for many years in the field of 3D visualization and graphics. Now, they also offer an extreme added value in the field of computer vision research. Numerous internationally renowned companies, research institutes and universities have already worked with Renderpeople datasets to feed machine learning models or simulations with detailed scanned human 3D data to work on difficult CV applications like 3D human perception and reconstruction or pose estimation.
Authentic high-resolution 3D scans help to generate ground truth synthetic training data for machine learning. A modern state-of-the-art 3D dataset fulfills the most important requirements for a successful ML-driven research approach: versatility, volume and quality. In addition, a 3D dataset offers various annotation possibilities for supervised learning that are not feasible with a conventional dataset of 2D imagery.
Feel free to get in touch with us. We are happy to provide you with individual services and advice for your specific concerns and needs.
Subscribe to the HumanDataset newsletter to be always up to date for new products and dataset extensions.
Posed | Single
Posed | Group
Rigged