Skip to main content

Installation Guide

Overview​

Syntonym Blur is a high-performance Snowflake Native Application designed to automatically anonymize personal data within visual media. By leveraging advanced computer vision, the application detects and blurs faces and vehicle license plates directly within your Snowflake environment.

This solution ensures distinct privacy compliance and data security by processing assets without them ever leaving the Snowflake governance boundary. It offers granular control, allowing users to adjust detection thresholds and selectively enable or disable blur types.

Key Features​

  • Automated Anonymization: Precision detection and blurring of faces and license plates.
  • Granular Control: Configurable sensitivity thresholds for detection logic.
  • Flexible Configuration: Selectively enable or disable specific blurring categories (e.g., blur faces but not plates).
  • GPU Acceleration: Optimized for high-throughput processing using Snowflake Compute Pools.
  • Secure Execution: 100% data residency; data never leaves the Snowflake security perimeter.
  • Broad Format Support:
  • Images: .jpg, .jpeg, .png
  • Video: .mp4, .avi

Installation & Setup​

1. Application Activation​

  1. Navigate to the Apps section in Snowsight.
  2. Locate SyntonymBlur and click Launch App.
  3. Select a Warehouse to drive the installation process and click Activate.
  4. Follow step 2.
  5. Note: After activation, please wait for the status to change to "System Online". This initialization process may take approximately 10–15 minutes.

2. Infrastructure Provisioning​

To utilize GPU acceleration for image and video processing, you must create a Compute Pool and grant usage to the application.

Execute the following SQL commands in a worksheet:

-- Create a Compute Pool with GPU support
-- Adjust MAX_NODES based on your throughput requirements
CREATE COMPUTE POOL IF NOT EXISTS app_pool
  MIN_NODES = 1
  MAX_NODES = 2
  INSTANCE_FAMILY = GPU_NV_2M;

-- Grant the application access to the pool
GRANT USAGE ON COMPUTE POOL app_pool TO APPLICATION SyntonymBlur;

-- Initialize the application services
CALL SyntonymBlur.app_schema.start_app();

> Reference: for a full list of supported GPU instance families, refer to the Snowflake Documentation: Create Compute Pool.

Usage Guide​

Data Ingestion​

You can ingest data via the Application UI or by copying files from an existing Snowflake Stage.

Option A: Upload via UI Use the application interface to directly upload supported image and video files.

Option B: Bulk Import via SQL To process existing datasets, copy files from your internal stage to the application's input stage.

Replace @my_stage with the name of your source stage.

COPY FILES INTO @SyntonymBlur.app_schema.input_stage
FROM @my_stage
PATTERN='.*\.(jpg|jpeg|png)$';

Accessing Results​

Processed (blurred) assets are automatically saved to the application's output stage.

To view the list of processed files:

LIST @SyntonymBlur.app_schema.output_stage;

Developer Information​

Syntonym AI Privacy-first computer vision solutions for secure analytics.

We enable organizations to utilize visual data for AI and analytics without compromising individual privacy.