Train U-Net deep learning model | Satellite Image Deep Learning

This video demonstrates how to train a deep neural network model with U-Net architecture for satellite image semantic segmentation tasks.
Check other videos in the series:
Part 1 - • Prepare input dataset ...
Part 2 - • Set up U-Net convoluti...
Part 3 - • Train U-Net deep learn...

[@eminatabeypeker6305]

thank u so much for the video and explanations. I have a question about the calculation methodology of accuracy; For example, it detected 60% of a polygon correctly, does the accuracy of this directly become 60%? Or if it is above a certain threshold value, is that image considered 100% accurate as binary?

[@geografif]

Hi, are you talking about the accuracy metric you got during the training phase?

[@eminatabeypeker6305]

@@geografif yes. actually i have a big dilemma. my testing label and prediction on test image commonly doesnt resemble however my test accuracy is higher than %80. i guess the reason of it is the images that do not contain any masks as i mentioned in previous video. i will try after i removed them.