Hi @DTS Engineer I have filed the request a long time ago. FB21757719 (Khmer Script Misidentified as Thai in Vision Framework)

While I understand Khmer (km-KH) is not currently supported, unsupported Khmer text is frequently being misclassified as Thai (th-TH) in features like Photos and Live Text. Could we adjust the detection confidence thresholds or fallback logic so unconfident matches don't default to Thai? Given the cultural sensitivity and distinct script differences, defaulting to "unsupported/unknown" would be a much better experience than misclassifying it.

While I understand Khmer (km-KH) is not currently supported, unsupported Khmer text is frequently being misclassified as Thai (th-TH) in features like Photos and Live Text. Could we adjust the detection confidence thresholds or fallback logic so unconfident matches don't default to Thai? Given the cultural sensitivity and distinct script differences, defaulting to "unsupported/unknown" would be a much better experience than misclassifying it. Thanks for your understanding and support!

Hi @DTS Engineer, There is no km-KH in the list. While I understand Khmer (km-KH) is not currently supported, unsupported Khmer text is frequently being misclassified as Thai (th-TH) in features like Photos and Live Text. Could we adjust the detection confidence thresholds or fallback logic so unconfident matches don't default to Thai? Given the cultural sensitivity and distinct script differences, defaulting to "unsupported/unknown" would be a much better experience than misclassifying it.