Fig. 8
From: SAD-Net: a full spectral self-attention detail enhancement network for single image dehazing
The generalization test results of SAD-Net on external datasets and the comparative experiments on the effects of convolution and attention on generalization test results. The first row shows the input images to the model, the second row shows the SAD-Net test results, followed by the test results of base_all, base_c, and base_a.
