Confidence scores different between Tensorflow model and TensorRT model

AI & Data Science Deep Learning (Training & Inference) TensorRT
1

Hi,

I’ve frozen my keras model to protobuf format, then converted it to uff. However, when i printed the classification confidence scores for the uff model, the scores seem to be significantly lower than the scores from the keras model. The input images are the same, and classification result is correct.

Is there a reason to the lower confidence scores from UFF model?

Thanks!