Multiscale Feature Filtering Network for Image Recognition System in Unmanned Aerial Vehicle

Table 1

The third and fourth columns show the architecture of MFFNet-50 and MFFNeXt-50, respectively.

Layers	Output size	MFFNet-50	MFFNeXt-50 (32 × 4d)
Convolution	112 × 112	Conv 7 × 7, 64, stride 2
Pooling	56 × 56	Max pool 3 × 3, stride 2
Stage 1	56 × 56
Stage 2	28 × 28
Stage 3	14 × 14
Stage 4	7 × 7
Classification layer	1 × 1	Global average pool
		1000 d fully connected, softmax