HSRD-100: 100 High-Quality 3D Human Scans Dataset from HumanScanRepository

Dataset Summary

HSRD-100 is a comprehensive 3D human scan dataset featuring 100 high-quality poses from 10 diverse individuals. This dataset provides a balanced representation of human demographics and poses, making it ideal for computer vision, machine learning, and 3D modeling applications.

Dataset Details

Dataset Statistics

• Total Poses: 100
• Unique Individuals: 10
• Poses per Person: 10
• Age Range: 21-64 years
• Gender Distribution: 50% Male, 50% Female
• Ethnicity Distribution:
  • Caucasian: 30%
  • African: 30%
  • Hispanic: 20%
  • Asian: 20%

Dataset Structure

Data Instances

Each data instance contains:

• 3D Model Files: Multiple LOD (Level of Detail) versions
  • {PoseID}-Scan.zip: High-resolution original scan at 5 mil polygons and 16k texture
  • {PoseID}-Scan-LOD0.zip: Highest quality mesh at 1 mil polygons and 16k texture
  • {PoseID}-Scan-LOD1.zip: Medium quality mesh at 100k polygons and 8k texture
  • {PoseID}-Scan-LOD2.zip: Lower quality mesh for performance at 10k polygons and 4k texture
• Source Files:
  • {PoseID}-RealityCapture.zip: Source photogrammetry data used for computation
  • {PoseID}-Photos.zip: Source photographs used for scanning rig with high-res 112 cameras
• Metadata Files:
  • person_meta.json: Individual demographic information. For more information Metadata Guide
  • pose_meta.json: Pose-specific metadata. For more information Metadata Guide
• Preview Images:
  • {PoseID}-Cover.jpg: Beauty image for dataset browsing (located in preview/ folder) with 4 sides of views of the scan.

Data Fields

• Person ID: Unique identifier for each individual
• Pose ID: Unique identifier for each pose
• Demographics: Age, sex, ethnicity, height, weight etc.
• Pose Information: Pose type, activity, clothing style etc.
• Technical Data: File paths, scan parameters, quality metrics

Data Splits

The dataset is organized by person and pose:

• Person-based organization: 10 individuals with 10 poses each
• Pose diversity: Various standing, walking, and activity poses
• Clothing variety: Casual, business, formal, and activewear

Dataset Creation

Curation Rationale

This dataset was created to provide a balanced representation of human diversity in 3D scanning applications. The selection criteria included:

• Demographic balance: Equal gender representation and diverse ethnicities
• Age diversity: Representation across adult age ranges
• Pose variety: Natural human poses and activities
• Clothing diversity: Various styles and contexts

Source Data

• 3D Scanning: Professional high-end 3D scanning equipment
• Photogrammetry: RealityCapture software for 3D reconstruction
• Quality Control: Consistent lighting and capture conditions
• Post-processing: Professional optimization and LOD generation

Preprocessing

• File Organization: Structured by Person ID and Pose ID
• LOD Generation: Multiple quality levels for different use cases
• Metadata Extraction: Comprehensive demographic and technical data
• Preview Generation: Beauty images for dataset browsing

Uses

Direct Use

This dataset can be used directly for:

• 3D Character Modeling: Reference poses for character creation
• Pose Estimation Training: Training data for pose recognition models
• Demographic Analysis: Studies on human representation in 3D data
• Animation Reference: Pose libraries for animation workflows

Downstream Use

Potential downstream applications include:

• Computer Vision: Human pose estimation, activity recognition
• Machine Learning: Training models for human behavior analysis
• Virtual Reality: Avatar creation and pose libraries
• Research: Human-computer interaction, accessibility studies

Out-of-Scope Use

This dataset should not be used for:

• Biometric identification: Not designed for person identification
• Medical diagnosis: Not intended for medical applications
• Surveillance: Not suitable for surveillance or tracking purposes

Data Considerations

Potential Biases

• Geographic bias: Scans may be limited to specific geographic regions
• Cultural bias: Clothing and pose choices may reflect specific cultural contexts
• Temporal bias: Data collected at a specific time period
• Equipment bias: All scans use the same scanning equipment and setup

Known Limitations

• Sample size: Limited to 100 poses from 10 individuals
• Pose variety: May not cover all possible human poses
• Demographic representation: While diverse, may not represent all populations
• Temporal factors: Clothing and styles may become dated over time

Ethical Considerations

• Consent: All participants provided informed consent for data collection
• Privacy: Personal identifiers have been removed from the dataset
• Representation: Efforts made to include diverse demographics
• Responsible use: Dataset intended for research and development purposes

Technical Specifications

3D Model Quality

• Format: OBJ (Wavefront OBJ)
• Average Polycount: 5,000,000 polygons per model
• Texture Format: PNG
• Texture Resolution: 16,384 pixels (4K+ resolution)
• File Structure: Organized by Person ID and Pose ID

File Organization

HSRD-100/
├── manifest/
│   ├── persons.jsonl          # Person metadata
│   ├── poses.jsonl            # Pose metadata
│   └── files.parquet          # File inventory
├── preview/                   # Thumbnail images
│   └── {PoseID}-Cover.jpg
└── data/                      # 3D model files
    └── {PersonID}/{PoseID}/
        ├── {PoseID}-Scan.zip
        ├── {PoseID}-Scan-LOD0.zip
        ├── {PoseID}-Scan-LOD1.zip
        ├── {PoseID}-Scan-LOD2.zip
        ├── {PoseID}-RealityCapture.zip
        └── {PoseID}-Photos.zip

Preview Images

• Location: preview/ folder (singular)
• Naming: {PoseID}-Cover.jpg (e.g., HSR0027-Body-006-Cover.jpg)
• Source: Beauty images from the original scan folders (Scan/ subfolder)
• Purpose: Thumbnail images for dataset browsing and preview
• Format: JPG images with high quality for visual identification

Citation

@dataset{HSRD-100_2025,
  title={HSRD-100: 100 High-Quality 3D Human Scans Dataset},
  author={Digital Reality Lab},
  year={2025},
  url={https://huggingface.co/datasets/digitalrealitylab/HSRD-100},
  license={cc-by-4.0}
}

License

This dataset is licensed under Creative Commons Attribution 4.0 International (CC BY 4.0).

What this means:

• ✅ Commercial use allowed - Companies can use this dataset in commercial products
• ✅ Personal use allowed - Individuals can use it for personal projects
• ✅ Modification allowed - Users can modify and adapt the dataset
• ✅ Distribution allowed - Users can share and redistribute the dataset
• 📝 Attribution required - Users must give appropriate credit and cite the dataset
• 🔗 Link to license - Users must provide a link to the license
• ⚠️ No additional restrictions - Users cannot apply additional restrictions

Contact

For questions about this dataset, please contact Digital Reality Lab or open an issue in the repository. If you are looking for larger dataset with 3d human scans and models check out our personal repository HumanScanRepository.

---

This dataset represents a carefully curated collection of 3D human scans designed to support diverse research and development needs in computer vision, machine learning, and 3D graphics.