"Multiple faces detected" during facematch between Selfie - ID

Problem Statement:

Facematch between ID and Selfie wrongly returning status 423 - "Multiple faces detected" error even though it's single ID and single person's face in the selfie.

Symptoms:

You may see 423 - "Multiple faces detected" error for the Facematch API when using a Smart Aadhaar card (as shown below). 

Cause and Impact:

1. We have delicately trained Face AI models for specific use cases because the quality of face image in a Selfie is different from the face image on an ID, this is how we get the best accuracy in the industry. 
2. This error occurs when the Facematch mode in the SDK code is set to selfie-selfie, but we are actually sending one selfie image and one ID image for comparison. 
3. The sample Smart Aadhaar image displays a smaller version of the document holder's image (as highlighted in red) on the top right side of the card, likely leading to the multiple face error.
4. Currently, this issue is only limited to Smart Aadhaar card (Any ID that has two face images on them).

Resolution: 

The simple lines of code that is being attached below should resolve this issue. This updates the facematch setting to selfie-id instead of selfie-selfie when facematch is being performed between a selfie and ID card.

Please find "HVNetworkHelper.makeFaceMatchCall" in your app where HyperVerge SDK is integrated and replace the code lines with below;

HVNetworkHelper.makeFaceMatchCall(context, faceMatchEndpoint, selfiePath, secondSelfiePath, HyperSnapParams.FaceMatchMode.FACE_ID, parameters, faceMatchHeaders, new APICompletionCallback() {})

If the issue persists even after making the changes, please try it out in a new sample app. If it works in the sample app, you may debug the differences between your main app and sample app to resolve the issue.

In case you have queries or need further assistance, feel free to write to us at support@hyperverge.co with the below details;

1. HyperVerge SDK version

2. If the issue persists in the sample app, please share it with us for debugging.

3. SDK integration code snippet

4. Screen recording of the issue and transactionId for the failed case.