/CNN_for_TEM_Segmentation

Python scripts for data set preparation and CNN training/inference

Primary LanguageJupyter Notebook

CNN_for_TEM_Segmentation

Python scripts for data set preparation and CNN training/inference

CNN_models: jupyter notebooks with examples of CNN model design and training procedures -HighRes - 4 convolutional steps with batch norm and second convolutional layer at each step (commented out) -Original - 3 layer CNN as used on 512x512 images -UNet_Leaky... - same architecture as HighRes, but with Leaky ReLU and learning rate scheduler

dataset_prep: -job.sh - bash script for parallel processing of images -label_training_set - example of applying filters, reconstruction to raw ETEM images -training_set_augmentation - example script for augmenting images

training_data: -sample_training_images/labels - set of 15 images/labels in a numpy array for processing by scripts in dataset prep ***Note: processing using job.sh, etc. uses jpg images; dataset_prep scripts can be easily modified to loop through a stack of images