Figure 6.

K-Means (k = 4) clustering on Nocodazole data subset (see Supplementary Material), using VGG13 and MoCov2 with different augmentations, aims to categorize cells’ morphological responses. Images nearest to each cluster’s centroid are based on Euclidean distance in representations. Clusters in (a), formed with color jitter, flips, rotation, affine transformation, and random cropping, focus on cell quantity per image.Clusters in (b), from rotations, center cropping, color jitter, and flips, consider specific phenotypes. Transformation details are in Supplementary Material.