Skip to main content
. 2019 Dec 14;19(24):5541. doi: 10.3390/s19245541

Figure 17.

Figure 17

SVHN: The proposed model based on 1 × 3 and 3 × 1 convolution pairs compared with a 3 × 3-based approach. Both variants are scaled to match in terms of FLOPs ranging from 1 to 10 million.