This repository is for the preprint|paper "Adaptive Partial Scanning Transmission Electron Microscopy with Reinforcement Learning", which presents a preliminary investigation into a new approach to compressed sensing. Sparse scans are piecewise adapted to specimens by a recurrent actor that learns to cooperate with a feedforward generator that completes scans.
Supplementary information is here.
Examples show test set 1/23.04 px coverage adaptive partial scans, target outputs and generated partial scan completions for 96x96 crops from STEM images.
To continue training the neural network; from scratch or to fine-tune it, you will need to adjust some of the variables at the top of train.py files. Specifically, variables indicating the location of your training data where to save logs and checkpoints.
Checkpoints for a fully trained model are here. They were saved after 500k and 1000k training iterations. To load the models, change the save location in the checkpoint file to your save location.
Datasets containing 19769 STEM images cropped or downsampled to 96x96 are here. Other dataset variants are also available.
The misc folder contains scripts to create graphs in the paper. In addition, read_loss_log.py can display loss logs output during training.
Jeffrey Ede: j.m.ede@warwick.ac.uk