webxos/3D_neural_network

3D Neural Network Dataset

Dataset Information

• Generated: 2026-01-09T00:01:04.850Z
• Session ID: session_1767916636309_sg9adv10z
• Total Steps: 1641
• Total Images: 1641
• Total Nodes: 153
• Total Connections: 303
• Session Duration: 0h 3m 48s

Dataset Structure

neural_network_3d_2026-01-09T00-00-59-568Z/
├── README.md
├── data/
│   ├── dataset.json              # Complete dataset (JSON)
│   ├── steps.csv                 # Time-series step data
│   ├── labels.csv               # Classification labels
│   ├── node_stats.csv           # Node statistics
│   └── connection_stats.csv     # Connection statistics
├── images/
│   ├── step_*.png              # High-resolution images
│   ├── thumb_*.jpg             # Thumbnails
│   └── *.png.meta.json         # Image metadata
└── metadata/
    ├── session_metadata.json   # Session information
    └── network_metadata.json   # Network structure

Time-Sequenced Data Format

The dataset features high-resolution PNG images (1641 in total) depicting each step of the network's growth, alongside CSV files for steps, labels, node stats, and connection stats. JSON files provide metadata, including session details and image specifics. Thumbnails in JPEG format (320x240) are included for quick previews. Classification labels cover aspects like complexity, topology, density, symmetry, and growth patterns, making it suitable for supervised learning tasks.

Potential applications extend to time-series prediction of network expansion, where models could forecast node additions based on historical steps; image-to-graph neural networks, converting visual representations into structured data; topology classification using provided labels; 3D visualization learning for rendering complex structures; and generative models for simulating network architectures. Its alignment by timestamps and FPS intervals ensures usability in dynamic, real-time ML tasks.

Potential Applications

It can support various ML workflows, such as developing image-to-graph models or studying 3D visualizations of neural networks. Users can load it via the Hugging Face Datasets library for integration into projects involving time-series analysis or graph generation.

Steps Data (CSV/JSON)

Each entry represents a precise moment in network construction:

• step: Sequential construction step
• timestamp: Unix timestamp (milliseconds)
• session_time_ms: Time from session start
• time_iso: ISO 8601 timestamp
• total_nodes: Current node count
• total_connections: Current connection count
• network_depth: Estimated network depth
• image_file: Associated image filename

Classification Labels

• complexity: Network complexity (simple/moderate/complex/very_complex)
• topology: Network topology (shallow/sparse/dense/deep/balanced)
• density: Connection density level
• symmetry: Spatial symmetry
• growth_pattern: Growth behavior

Node & Connection Statistics

Detailed statistics for each construction step including:

• Node type distribution
• Connection type distribution
• Average weights and layers
• Activation function usage

Image Data

• Resolution: Native canvas resolution
• Format: PNG (lossless)
• Timestamps: Precisely aligned with tabular data
• Metadata: Includes classification labels
• Thumbnails: 320x240 JPEG

Multimodal Data Alignment

All data streams are precisely synchronized:

1. Tabular data captured at exact timestamps
2. Images captured at specified FPS intervals
3. Classifications computed for each step
4. Metadata preserved for each image

Usage Examples

• Time-series prediction of network growth
• Image-to-graph neural networks
• Classification of network topologies
• 3D visualization learning
• Graph generation models

Citation

@dataset{3d_neural_network,
  title = {3D Neural Network Dataset},
  author = {webXOS},
  year = {2026},
  url = {webxos.netlify.app}
}