Fig. 1: An example of denoising X-ray diffraction data using a deep CNN.
From: Weak signal extraction enabled by deep neural network denoising of diffraction data
a,b, A real experimental LC frame (exposure time 1 s) (a) is used as an input to a deep CNN (b) trained to remove the noise. c, The denoised output reveals a CDW signal (red), barely visible in the raw LC data. d, The real experimental HC frame (exposure time 20 s) for comparison. e, A stack of denoised X-ray intensity frames as in c. Arrows indicate the projected reciprocal coordinates Q = (h, k, ℓ). f–h, One-dimensional projected scans through Q ≈ (0.23, 0, 8.5) along the h (f), k (g) and ℓ (h) reciprocal space axes, in units of r.l.u. For each projected scan, a background subtraction has been performed (see main text). Gaussian fits for HC and denoised output profiles are indicated by solid red lines. The data points depicted in the denoised output profile are computed as the mean value over five training runs of the IRUNet neural network with different initial conditions. Error bars for LC and HC are shown under the assumption of counting statistics. Error bars for the denoised output are shown as the s.d. over the mentioned training runs. The clock symbols indicate relative counting time, and the network symbol indicates the denoised LC produced by the neural network.
