Can we extract confidence score of all classification in maskRCNN?

[Jeff-Gui]

Hi, I'm working with semantic segmentation on time-lapse data. Therefore, it will be helpful to smooth classification using a temporal sliding window. If my understanding is correct, the 'score' field under prediction return dict refers to the confidence of classification. I wonder if we can extract all confidences of all classes. Thanks~