Fig. 2: Comparisons among BCNN, CNN, End-To-End BCNN and CNN in the simulated SPI trained with the MNIST database.
From: Approximating the uncertainty of deep learning reconstruction predictions in single-pixel imaging
a A representative ground-truth image in the testing dataset, input images to the BCNN and CNN calculated from the LSQR-approximated inverse model matrix, 1D raw measurement data as the input to End-To-End BCNN and End-To-End CNN, and the predictions from BCNN, CNN, End-To-End BCNN and End-To-End CNN at the 8×, 16×, 32× and 64× compression ratios. b The MAEs of the predicted images in BCNN, CNN, End-To-End BCNN and End-To-End CNN at the four compression ratios. c The SSIMs of the predicted images in BCNN, CNN, End-To-End BCNN and End-To-End CNN at the four compression ratios. d Averaged pixel values of the predicted model uncertainties in BCNN and End-To-End BCNN in the testing dataset at the four compression ratios. e Averaged pixel values of the predicted data uncertainties in BCNN and End-To-End BCNN in the testing dataset at the four compression ratios. The error bars represent the standard deviation of the corresponding parameters from 100 testing images.
