基于孪生残差神经网络的GF-2影像林地变化检测

艾遒一,黄华国,郭颖,刘炳杰,陈树新,田昕

Forest Change Detection based on Siamese Neural Network with GF-2 Image: A Case of Jiande Forest Farm, Zhejiang

Qiuyi AI,Huaguo HUANG,Ying GUO,Bingjie LIU,Shuxin CHEN,Xin TIAN

表2 ResNet50网络组成

Table 2 ResNet50 network structure

分层名称	输出尺寸	50-layer
conv1	112×112	7×7,64,stride 2
conv2-v	56×56	3×3 max pool,stride 2
conv2-v	56×56	$[\begin{matrix} 1 \times 1,64 \\ 3 \times 3,64 \\ 1 \times 1,256 \end{matrix}] \times 3$
conv3-x	28×28	$[\begin{matrix} 1 \times 1,128 \\ 3 \times 3,128 \\ 1 \times 1,512 \end{matrix}] \times 4$
conv4-x	14×14	$[\begin{matrix} 1 \times 1,256 \\ 3 \times 3,256 \\ 1 \times 1,1024 \end{matrix}] \times 6$
conv5-x	7×7	$[\begin{matrix} 1 \times 1,512 \\ 3 \times 3,512 \\ 1 \times 1,2048 \end{matrix}] \times 3$
	1×1	average pool 1 000-d fc,softmax
FLOPs		3.8×10⁹