http://www.vlfeat.org/matconvnet/
QuickStart 下载 MatConvnet
eval.m --> Layer.m
注释掉 %clear inputs if not needed anymore 模块,保存每一层的输出
提取 penultimate layer (pool 5) 1024维features;
features = squeeze(net.vars(net.getVarIndex('cls3_pool')).value) ;
http://www.vlfeat.org/matconvnet/
QuickStart 下载 MatConvnet
eval.m --> Layer.m
注释掉 %clear inputs if not needed anymore 模块,保存每一层的输出
提取 penultimate layer (pool 5) 1024维features;
features = squeeze(net.vars(net.getVarIndex('cls3_pool')).value) ;