Release Note - ID Forgery Detection [INDO] V2.1

Product Release

Product name: ID Forgery Detection [INDO] V2.1

Release date: 18 May 2023

Summary

In this minor release 2.1, we had two key changes:

• Improved the retake and screenshot predictive performance.
• Resolved bug where previously an image was determined to be forged ("failed"), but no reasons were provided.

Details

1. Retake and Screenshot performance improvement

We released new and updated models to tackle the retake and screenshot forgery types, and based on our benchmark dataset, the false acceptance rate (FAR), decreased by 5% and 2% respectively.

2. Fixed bug of ID image failed but no reasons were provided

We identified a bug where ID cards failed but no reasons were provided in the detail field. Based on our evaluation, these issues were due to our handling of the poor-quality module which we have promptly fixed, and it will now return "Poor Quality" for relevant cases.

Example (V2.0) - No reasons provided

{
"code":"SUCCESS",
"message":"OK",
"data":
 {
 "result":"fail",
 "detail":[]
 },
"extra":null,
"pricingStrategy":"PAY"
}

Example (V2.1) - "Poor Quality" bug fixed

{
"code":"SUCCESS",
"message":"OK",
"data":
{
"result":"fail",
"detail":[Poor Quality]
},
"extra":null,
"pricingStrategy":"PAY"
}