search

heyufan1995 / oct_preprocess

OCT preprocessing: Fully Convolutional Boundary Regression for Retina OCT Segmentation
44 stars 10 forks source link

The purpose of oct_preprocess #2

Closed zhilingxuQsQ closed 1 year ago

zhilingxuQsQ commented 2 years ago

Why we have to flatten the BM/RPE before we do the following segmentation task or other tasks? Whether it is necessary to improve the accuracy of retina segmentation?

heyufan1995 commented 1 year ago

The initial purpose is to reduce the GPU memory, because the B-scans are too large. I feel like it's okay to skip the flattening if GPU memory is enough nowadays.