Overview

Autonomous driving datasets often contain sensitive information, such as faces and car license plates, which need to be protected for privacy compliance. Our de-identification model API is designed to accurately detect and anonymize faces and car license plates, ensuring that the data is safe and compliant with regulations such as GDPR and CCPA. This guide explains how to use the API for de-identification in autonomous driving datasets.

Purpose of the De-identification Model

The primary goal of the de-identification model is to protect personal privacy by automating the anonymization process of identifiable elements within autonomous driving datasets. Key functionalities include:

Face Detection and Anonymization: The model accurately detects faces in images or videos and applies anonymization techniques like blurring or masking to ensure the individual cannot be identified.
License Plate Detection and Anonymization: Car license plates are detected and anonymized using similar techniques, ensuring that personal information related to the vehicle owner is safeguarded.

How to Use the API

API Call Structure
The de-identification model API operates via a RESTful interface. You can upload images or videos from your dataset, and the API will process them, returning anonymized results. Key API details include

API Endpoint: Bounding Box

curl --request POST \
     --url https://api.models.aimmo-mlops.com/v1/de_identification_bbox\
     --header 'accept: application/json' \
     --header 'content-type: application/json' \
     --data '{"threshold":0.5}'

curl --request POST \
     --url https://api.models.aimmo-mlops.com/v1/de_identification_polygon \
     --header 'accept: application/json' \
     --header 'content-type: application/json' \
     --data '{"threshold":0.5}'

curl --request POST \
     --url https://api.models.aimmo-mlops.com/v1/de_identification_polygon \
     --header 'accept: application/json' \
     --header 'content-type: application/json' \
     --data '{"threshold":0.5}'

curl --request POST \
     --url https://api.models.aimmo-mlops.com/v1/de_identification_polygon \
     --header 'accept: application/json' \
     --header 'content-type: application/json' \
     --data '{"threshold":0.5}'

API Endpoint: Polygon

curl --request POST \
     --url https://api.models.aimmo-mlops.com/v1/de_identification_polygon \
     --header 'accept: application/json' \
     --header 'content-type: application/json' \
     --data '{"threshold":0.5}'

Request Type: Post
Request Parameters: Image or video files in the request body

Headers:

headers = {
    "accept": "application/json",
    "content-type": "application/json",
    "authorization": "Bearer {token}"

Authorization: Bearer token with your API key
Content-Type: multipart/form-data

De-identification Processing Options

The API offers several de-identification techniques to handle sensitive information

Blur: Faces and license plates are blurred to prevent identification.
Mask: Faces and license plates are covered with a solid color mask, making them completely unrecognizable.

Use Cases

Dataset Preparation: The API is perfect for preparing autonomous driving datasets for machine learning training by anonymizing personal data without affecting data quality.
Research and Demonstrations: During research presentations or public demonstrations, the anonymization ensures compliance with privacy laws.
Privacy Protection: Protects sensitive data in public videos while ensuring legal compliance and minimizing the risk of data misuse.