3D segmentation is LITERALLY SO COOL!!! I already did a 2D segmentation project, so projects within this space would be the perfect challenge because I have never worked with 3D data + the preprocessing, model, training/validation/testing, and visualization are so different.
I looked into some datasets and I decided that I want to work with an electron microscopy dataset, which has zoomed in 3D models of whole cells and their organelles. My 3D segmentation model will hopefully be able to segment the 7 different classes in each image: background, cell, mitochondrion, canalicular channel, alpha granule, dense granule, and dense granule core.
I also came across the BraTS Challenge because alongside researchers, I'm building a model to segment out tumors from 3D brain scans (stay tuned for more information next month 👀)