Figure 2
From: Learning deep features for dead and living breast cancer cell classification without staining
(a) Bright-field and corresponding propidium iodide fluorescence images. The columns under Live patches (green) show images with mostly live cells. The columns under Dead patches (red) present images with mostly dead cells in our dataset (images rendered using Matplotlib 3.3.3 https://matplotlib.org/). (b) Left, samples of empty and non-empty patches with the associated barcode visualization of the 512-dimensional feature vector from ResNet-18 last convolutional output per image. Right, dimensionality reduction visualization of patch features and the high-dimensional decision function (green level-sets) learned by the SVM. Notice how empty patches mostly lie inside the highlighted region (images rendered using Seaborn 0.11.1 https://seaborn.pydata.org/ and Matplotlib 3.3.3 https://matplotlib.org/).
