Figure 1: simulation of ∼0.04 deg2 region of GOODS-North by Coogan et al. (2023)
Figure 2: zoom of previous figure Coogan et al. (2023)
Figure 3: Mandelbaum et al. (2014)
Figure 4: IllustrisTNG simulations
Our goal:
"Infuse deep learning map-to-map translation models
with the physical model from simulations."
Benchmark of generative models we're investigating:
Figure 5: pix2pix scheme
Figure 6: U-Net following Ronneberger et al. (2015)
"Attention is (almost) all you need!"
Figure 7: Input
Figure 8: Output (pix2pix with Attention U-Net)
Figure 9: Ground truth
Figure 10: Input
Figure 11: Output (pix2pix with Attention U-Net)
Figure 12: Ground truth
Figure 13: Input
Figure 14: Output (pix2pix with Attention U-Net)
Figure 15: Ground truth
Figure 16: Input
Figure 17: Output (pix2pix with Attention U-Net)
Figure 18: Ground truth
Figure 19: Input
Figure 20: Output (pix2pix with Attention U-Net)
Figure 21: Ground truth
Figure 22: Input
Figure 23: Output (standard DDPM)
Figure 24: Ground truth
Figure 25: model using pix2pix+Attention
Figure 26: data
Figure 27: Input
Figure 28: Output (pix2pix with Attention U-Net)
Figure 29: Ground truth
Figure 30: Input
Figure 31: Output (pix2pix with Attention U-Net)
Figure 32: Ground truth
Figure 33: Input
Figure 34: Output (pix2pix with Attention U-Net)
Figure 35: Ground truth
Email: philipp.denzel@zhaw.ch