Fig. 2: Hologram filtering neural network. | Nature Communications

Fig. 2: Hologram filtering neural network.

From: Deep learning-based incoherent holographic camera enabling acquisition of real-world holograms for holographic streaming system

Fig. 2

a Neural network architecture for filtering holograms. The network contains one strided convolution block, nine residual blocks and one transposed convolution block. b Training procedure of the neural network. The real and imaginary parts of the holograms are presented only for the green channel. GP GP lens, P linear polarizer, dc distance between the central plane and GP, di distance between the target image and GP, ASM angular spectrum method, d-ASM depth-corrected ASM. c, d Training dataset examples. c Reference target images displayed on a 2D display at various depth positions and d the corresponding captured holograms.

Back to article page