gvacaliuc
commented
6 years ago need to formalize color transfer dialogue and present examples -- we're really just performing PCA, so it might be interesting to see some 3D plots on nuclei data. I'd say we can present PCA plots for clarity as well as support for our system of equations, while we should save the nuclei painting for results
need to discuss and present examples of our hand designed features. I'll make a large figure with our transformations applied to simulated data. The data should have a fair degree of noise with several discs w/ random affine transformations applied.
need to explain what sgd regressor and pa regressor solve mathematically and their respective hps.
need a figure to present the unet.
need to explain the watershed process as well as our non-max supression.
if we have enough time i'd like to look into doing the non-maximum suppression with a network. basically it would be mask-rcnn except we'd cut out the rpn because our nuclei centers are very very good, we just occasionally have too many. we're ~~probably definitely not going to have time in the next day (since poster is due in 28 hours) to do this.~~
planning doc