Fig. 1: Training a deep neural network for coherent image reconstruction. | Light: Science & Applications

Fig. 1: Training a deep neural network for coherent image reconstruction.

From: Deep learning in holography and coherent imaging

Fig. 1

a Training workflow of an image reconstruction deep neural network, including data acquisition, physics-based constraints, image registration, and training. b Typical network training and testing procedure, where the network learns to match the input image to a target label (ground-truth image) using a given loss/cost function

Back to article page