Computer Vision News - February 2022

36 Medical Imaging Tools Application of fmi Library to fastai in Medical Imaging and Deep Learning I hope everyone in those parts of the world that changes the year, had a great celebration for the new year! I also hope that everyone had a wonderful time in this quite hard winter, hoping for many new, bright days to come... Thismonthwewill investigate the applicationof the fmi library to the existing, great fastai approach to medical imaging and deep learning . As you would probably remember from the earliest articles, fastai has created a user-friendly deep learning library, that can be used with almost zero overhead, and has great documentation. We are going to improve this experience using the fmi library! Keep reading to find out and maybe get a few innovative ideas for your future projects! Let’s start with basic image processing, imports for the libraries needed and check the system information. Important to have aGPU to run those examples! You can always find a freeGPU instance using Google’s Collab or other similar platforms  import pydicom,kornia,skimage from fastai.vision.all import * from fastai.medical.imaging import * from torchvision.utils import save_image import seaborn as sns from fmi.pipeline import * from fmi.explore import * from fmi.preprocessing import * from fmi.examine import * from fmi.train import * from sklearn.model_selection import train_test_split import timm matplotlib.rcParams['image.cmap'] = 'viridis' import warnings warnings.filterwarnings("ignore", category=DeprecationWarning) system_info conveniently lists the current fastai , fastcore versions as well as cuda , pydicom and kornia versions syst em_info() fastai Version: 2.4