False Positive Face Detection in .tflite Model – Real Faces Misclassified as Spoof

[Rohan-ManekTech]

I have integrated this .tflite model into my Flutter project for face detection, but I am experiencing an issue where real faces are frequently misclassified as spoofed faces. In many cases, the model returns a confidence score of 0.9 for spoofing on real faces.

Could you please: 1. Update the .tflite model to improve spoof detection accuracy? 2. Provide steps or guidance on how to properly train the model to reduce false positives?

Any help on improving the model’s accuracy would be greatly appreciated. Thanks!