Skip to main content
. 2020 Aug 21;2020:1375957. doi: 10.34133/2020/1375957

Figure 4.

Figure 4

Overview of SFC2Net. The MixNet-L backbone first extracts feature maps that are further fused by multilayer fusion modules (MFM). Then, the redundant module processes multiscale feature maps to generate a redundant class map. Finally, after inverse quantization and deredundancy, SFC2Net outputs the count map. The final count of the input image is computed by summing each pixel in the count map.