1.模型就用程序自帶的caffenet模型，位置在 models/bvlc_reference_caffenet/文件夾下, 將需要的兩個配置文件，復制到myfile文件夾內
2. 修改solver.prototxt（用notepad）
net: "examples/myfile/train_val.prototxt" #
test_iter: 2
test_interval: 50
base_lr: 0.001
lr_policy: "step"
gamma: 0.1
stepsize: 100
display: 20
max_iter: 500
momentum: 0.9
weight_decay: 0.005
solver_mode: CPU #我用的是cpu所以改成了cpu
3.修改train_val.prototxt
name: "CaffeNet"
layer {
name: "data"
type: "Data"
top: "data"
top: "label"
include {
phase: TRAIN
}
transform_param {
mirror: true
crop_size: 227
mean_file: "examples/myfile/mean.binaryproto"
}
# mean pixel / channel-wise mean instead of mean image
# transform_param {
# crop_size: 227
# mean_value: 104
# mean_value: 117
# mean_value: 123
# mirror: true
# }
data_param {
source: "examples/myfile/train_db"
batch_size: 256
backend: LMDB
}
}
layer {
name: "data"
type: "Data"
top: "data"
top: "label"
include {
phase: TEST
}
transform_param {
mirror: false
crop_size: 227
mean_file: "examples/myfile/mean.binaryproto"
}
# mean pixel / channel-wise mean instead of mean image
# transform_param {
# crop_size: 227
# mean_value: 104
# mean_value: 117
# mean_value: 123
# mirror: false
# }
data_param {
source: "examples/myfile/test_db"
batch_size: 50
backend: LMDB
}
}
其他不需要改變。
4.訓練和測試（具體見langb2014的http://blog.csdn.net/langb2014/article/details/50458014）
（1）運行.sh
cygwin到caffe根目錄下輸入
bin/caffe train -solver examples/myfile/solver.prototxt
解釋：
【1】bin/caffe 你的caffe.exe所在目錄
以下就不說了看langb2014大神的特別清楚！！！
結果（我的還沒有訓練完，因為我的是只有cpu運行的超級慢）：
看大神的結果：
到現在，成功運行了一個caffe例子
（2）運行.bat
新建run_test.bat，內容為
D:/deeptools/caffe-windows-master/bin/caffe.exe train --solver=D:/deeptools/caffe-windows-master/examples/myfile/wintest/solver.prototxt
pause
然后solver.prototxt和train_val.prototxt都改成了絕對路徑，其他的沒變