U-Net — all tests passed with grade 66/100


For the Image Segmentation with U-Net assignment I have passed all the tests and got reasonable outputs. But the grader gives 66/100 and says “Code Cell UNQ_C3: Function ‘unet_model’ is incorrect”. Could anyone tell what is wrong with my code implementation?

Many thanks,

Are you still having an issue with this exercise?

Be sure you specify the number of filters in every conv_block() and upsampling_block().

I’ve got the same problem. I checked the number of filters in all blocks. Used double in conv_blocks and half in upsampling blocks. My code seems correct and I passed all tests, but still got 66/100.

{I have edited my reply by deleting the previous text}

most probably inside the u-net function you put an actual number instead of the variable filter_num which you have to multiply or divide step-wise based on whether you are expanding or contracting the tensor


Yes, this was my fault, thank you @TMosh, @jgeng