#imagereconstruction

[kiranholographicsblog]

best automatic hologram applicator

[throughmachineanddarkness]

Fig. 4 Selection of intermediate feature maps in the first (top row) and second (bottom row) layer of the neural network.

‘One of the strengths of CNN is that the feature map at each layer is an image that can be visualized. We report in Figure 4 a selection of features maps from the output of the first and second layers. We can see that the first layer contains some low pass or directional filtering whereas layer two contains more semantic information.

Astronomical image reconstruction with Convolutional Neural Networks, Rémy Flamary, Université Cote d’Azur, Nice, France. Full paper available here.

[throughmachineanddarkness]

Fig. 3. Visual comparison of the original and AI reconstructed images for M51a. 

Astronomical image reconstruction with Convolutional Neural Networks, Rémy Flamary, Université Cote d’Azur, Nice, France. Full paper available here.