Semantic segmenation task Score as 0

Hello ,Im working on the kaggle semantic segmentation challenge,The testing is done the predictions are out, but when i make the submission i get a score of 0.I get decent predictions,i was looking forward to a less score but i ended up getting a zero,Im not sure what the mistake here,The predictions have to be converted to run length encoding before submission,

some of my predictions,

and the run length code look like this

If anyone has any suggestions kindly let me know.
Thanks in advance

I’m not sure what prob_to_rles does, but do you applz RLE on each cell?
If you just submit the RLE mask images, you will get a score of 0, since this competition requests each cell to be in a separate row AFAIK.
The Kaggle forum should be a better platform for this question though. :wink:

Hi @ptrblck

Sorry for the late reply,I was sick and couldnt show up for work.

Im just taking the array of predictions and passing it to the prob_to_rles and then doing the submission.
I guess im going wrong somewhere in there!!

Not many pytorch notebooks to follow in kaggle,So this is mostly the default place for any help,

I hope you are feeling better!
So “array of predictions” means each row contains a cell mask?
If not, I think that’s what this competition wants. Could you try this approach and let me know if it helps?

Thanks ,much better.

yes,i think each row contains a cell mask,Let me recheck the process again.


By array of predictions, i mean , if i do predictions[0],I get

array([[ 0.,  0.,  0., ...,  0.,  0.,  0.],
   [ 0.,  0.,  0., ...,  0.,  0.,  0.],
   [ 0.,  0.,  0., ...,  0.,  0.,  0.],
   [ 0.,  0.,  0., ...,  0.,  0.,  0.],
   [ 0.,  0.,  0., ...,  0.,  0.,  0.],
   [ 0.,  0.,  0., ...,  0.,  0.,  0.]], dtype=float32)

Now this i feed into the prob_to_rles function,and then it does its encoding.I mean the main predictions array which contains the 65 images.
Is this Okay?

If the array is for mask for one cell in the image, it should be ok.
However, if this array is the mask for all cells in the image, you will probably get a score of 0.

Oh Okay.Let me look into that,Thanks for your time.