search

browatbn2 / VLight

MIT License
5 stars 2 forks source link

"Scale" is not clear in the paper. #1

Closed Limingxing00 closed 3 years ago

Limingxing00 commented 3 years ago

Hi,

Today I had the honour to read the paper, "The Unreasonable Effectiveness of Encoder-Decoder Networks for Retinal Vessel Segmentation".

Could you explain the "Scale" in Table 1 of the paper? What's the meaning of " 2/2" , " [2-4] / 4" and " [2-4] / 2+3+4"? I cannot find the detail in the paper.

Best

browatbn2 commented 3 years ago

Hi, thanks for your interest and reading the paper!

x/y refers to the image scale for training/test. So 2/2 is two times the image size for training and test. [2-4] / 4 would mean training on a continuous range between 2x and 4x with testing on 4x image size. 2+3+4 means testing was done on scales 2, 3, and 4 and the outputs combined (averaging).

Limingxing00 commented 3 years ago

Thank you for taking the time to answer the questions.

So when the patch size is 512x512 and the scale is 2, the input size is 256x256?

browatbn2 commented 3 years ago

Patch size and input size are always the same. Scale 2 would double the image size, thus increase the number of patches by a factor of 4.

Limingxing00 commented 3 years ago

Thanks!