how can I evaluate the segmentation results quantificationally?

[Rongrong-LIU525]

Hello, thank you again for sharing your work. You have provided a notebook to see the prediction example , and I want to know how you evaluate the results quantificationally. You extract the resulting segmentation picture from the ipynb, then calculate the F-measures and other metrics somewhere else? Thank you.

[Rongrong-LIU525]

I want to extract the prediction result and measure it with my former matlab programmes. But when I download directly or save the segmentation result, the pixels are changed. I have tried but still can not save it with the original shape. Do you have any idea?

[lim-anggun]

Hi @MaggieGlory , You don't need to calculate the score using Matlab since scikit learn provides useful functions to compute the score. What you need to do is:

1. Extract foreground mask (predicted values from sigmoid function are between 0 and 1.) e.g. probs

2. Binarize the extracted mask in step 1, say:

   threshold = 0.8
   probs[probs<threshold] = 0.
   probs[probs>=threshold] = 1.

3. Compute the F1-score using Scikit Learn

I will upload my code on how to compute F1-score later this week. Stay tuned!

[Rongrong-LIU525]

Thanks so much for your reply. -