akhilesh64 / image-segmentation-u-2-net Goto Github PK

I trained for 100 epochs as a test on the train data from DUTS using the duts train script but I got an error after running the predict file

ValueError: Mismatch between expected batch size and model output batch size. Output shape = (7, 1, 256, 256, 1), expected output shape = shape (1, 1, 256, 256, 1)

using tf 2.1

What is threshold for?

I am not seeing any differences by modifying it

Dear sir
Thank you for this simple and great implementation of the u2net architecture
I work in a cell instance segmentation project . I have prepared my data in two folders, the first one is the cell images the second one is the mask PNG images created from rle annotations
I have run the training for 1000 epochs but when I want to predict on test image I got black image I can not see the mask

I am trying to train using DUT dataset. after I ran code U2-net.py nothing gets output. I am expecting output of learning, loss and current number of iterations. Any idea? thanks

Hi, I have a dataset of about 30,000 images for training. It take for ever to load them. Is there other way which is faster to load this amount of data? thanks