AI Recognition Accuracy Response Program
Multi-dimensional validation schemes are recommended for possible misrecognition by the Google Vision API:
- Cross-validation tools::
- Comparative analysis was also performed using Microsoft Azure Computer Vision
- Local deployment of the open source Detectron2 model for secondary validation
- Points for Manual Inspection::
- Note the AI labeled confidence scores (professional tools show probability values)
- Focus on verifying key elements such as physical features and textual information
- Be wary of vague expressions such as 'possible' and 'suspected'.
- Active jamming technology::
- Adding visual noise to sensitive photos (recommended intensity 5-10%)
- Fine-tuning facial features using GAN generative adversarial networks
- Interfering with AI Recognition Using Adversarial Sample Techniques
Establish a three-stage workflow of 『recognition-verification-processing』, and when AI recognizes sensitive information, manually confirm it before deciding how to handle it. Business users can consider building an automated audit pipeline.
This answer comes from the articleThey See Your Photos: Analyzing Photo Privacy Information Based on Google VisionThe
May not be reproduced without permission:AI productivity tools " How to overcome the problem of privacy misclassification due to AI misidentification?
