Predict the age on an 'average' image

[Dyex719]

The output varies too quickly to be interpreted well at times. An easy solution would be to take a 100 frames and print the average value of the predictions, but I was wondering how this would do against taking an average embedding of, say 100 frames, and run the prediction on that average image.

I try to get the embeddings of the images by storing the outputs of the final dense layer, however this slows down the system considerably. I notice that results = self.model.predict(face_imgs) however doesn't do so.

Is there a better way to achieve this?

[roiksail]

I have the same problem,I want to use the average of predictions in several frames.how could i do it?