caffe_ssd-matlab调试

models:
－solver.prototxt: 各种参数(包括train.prototxt的路径)

train_net: “models/Context_Net/train.prototxt”
base_lr: 0.0005
display: 10
max_iter: 60000
lr_policy: “step”
gamma: 0.1
momentum: 0.9
weight_decay: 0.0005
stepsize: 40000
snapshot: 40000
snapshot_prefix: “models/SSD_FACE_300x300”
solver_mode: GPU
device_id: 0
debug_info: false
snapshot_after_train: true
test_initialization: false
average_loss: 10
iter_size: 2
type: “SGD”
eval_type: “detection”
ap_version: “11point”

－train.prototxt: 训练各层设置

需要修改最底层的conv4_3…conf的output的个数为６＊类数
最底层的其他…conf的output的个数为３＊类数

－VGG_ILSVRC_16_layers_fc_reduced.caffemodel

functions
-+Net/net_conf.m

修改num_classes类数，数据路径