Intended Use

Saifr's image detection model identifies images that may be noncompliant based on regulatory norms. It has been finetuned with a vast database of images (labelled by subject matter experts). This data includes both compliant and noncompliant examples across the financial sector, enabling it to recognize and flag potentially noncompliant image content. The input for the model is an image, and output would be whether the image is likely to be compliant or not. If it is flagged for likely noncompliance, the model provides the relevant category.
Currently, the model primarily focuses on the noncompliant objects in the images. Future versions plan to include text present in the image as a factor in assessing noncompliance.

License

The use of this model is subject to the Saifr License Agreement available at https://saifr.ai/azure-custom-license Use of Saifr’s models is subject in all respects to the custom license agreement between Saifr and the end user.  Saifr models are not intended to replace the end user’s legal, compliance, business, or other functions, or to satisfy any legal or regulatory obligations.  Note that all compliance responsibilities remain solely those of the end user and that certain communications may require review and approval by properly licensed individuals.  Saifr is not responsible for determining compliance with rules and will not be liable for actions taken or not taken based on model use. 

Sample Inputs and Outputs (REST)

Example Request

You can use cURL or any REST Client to send request. Just add your token and test.

    curl https://url.com -X POST -d '{ "images" : [{"id": string, "image_object": string }]}' -H "Authorization: Bearer <TOKEN>" -H "Content-Type: application/json"

Sample Input

An array of images in the following format

{
  "images": [
    {
      "id": string,
      "image_object": string
    }
  ]
}

where -

• id: Unique ID assigned to the image (Required)
• image_object: The base64 representation of image.

Supports the following image formats - JPEG / TIFF / Bitmap / GIF / PNG

Sample Output

An array of imageClassification objects

{
  "image_classifications": [
    {
      "id": string,
      "category": string,
      "is_compliant": bool
    }
  ]
}

where -

• id: The unique ID assigned to the image as part of input
• is_compliant: Indicates whether the image is compliant or not
  • True - indicates image is compliant
  • False - indicates image is not compliant
• category: The category of the image. Possible values are -
  • 'money tree'
  • 'mansions'
  • 'fancy cars'
  • 'fancy watches'
  • 'yachts'
  • 'gulf stream jets'
  • 'bags of money'
  • 'bank vaults'
  • 'treasure chest'
  • 'expensive jewelry and brands'
  • 'stack of money coins chips'
  • 'gold bar'
  • 'charts arrows going up'
  • 'pot of gold'
  • 'golden nest egg'
  • 'piggy bank'
  • 'american flag' - US Flag Code prohibits use of the American flag for advertising purposes
  • 'negatives' - Indicates the image is compliant and does not belong to any of the above categories