BWS-Data-Solutions/BWS-Privacy-Blurred-POC

Privacy-Native Blurred Environments (POC)

High-quality, heavily blurred environmental images designed specifically for computer vision and depth estimation training — built from the ground up to eliminate biometric, PII, and Right of Publicity risks.

This Proof of Concept dataset is 100% proprietary (shot and owned by Brett Williams Studios). No scraped, web-crawled, or third-party data.

Why This Dataset Exists

With the CLEAR Act 2026 disclosure requirements now active, many teams need training data that is verifiably clean. This collection removes facial geometries, identifiable features, and personal data while preserving rich environmental context perfect for:

• Depth estimation & monocular depth models
• Bokeh simulation & portrait effects
• Background composition & segmentation
• Scene understanding
• Synthetic data augmentation

Dataset Properties

• Total Images (POC): 120 high-fidelity samples (full 25k+ production set available)
• Style: Multi-environment indoor/outdoor scenes with intentionally blurred human figures
• Resolution: High-resolution originals (downsampled versions also available)
• Format: JPG + metadata
• IP Status: Fully cleared for commercial AI training use

Dataset Splits

Split	Images	Purpose
train	96	Main training / development
validation	24	Evaluation & quick preview

Legal & Compliance

• Biometric/PII free by design
• Delivered with SHA-256 cryptographic manifests for audit trails
• Non-exclusive commercial licensing available
• Full production dataset available via secure AWS S3 -https://www.kaggle.com/datasets/createphotos/rivacy-native-blurred-environments-poc

Getting Started

from datasets import load_dataset

dataset = load_dataset("BWS-Data-Solutions/BWS-Privacy-Blurred-POC")

# View an example
example = dataset["train"][0]
example["image"].show()