Caffe的全称是Convolutinal Architecture for Fast Feature Embedding。核心开发语言是C++和CUDA，支持命令行、Python和MATLAB接口。支持CPU和GPU执行，支持单机多卡并行，不支持多机并行。

依赖库 matlab opencv python - anaconda

boost gflags glog leveldb/lmdb protobuf hdf5 snappy

MKL >= OpenBlas > AtLas

代码分析

总体结构

Caffe使用Protobuf文本格式（plaintext protocol buffer schema）定义Solver、Net、Layer。描述这些组件的文本格式的定义都放在caffe.proto文件中。

操作接口

http://caffe.berkeleyvision.org/tutorial/interfaces.html

命令行操作

Training
# train LeNet
caffe train -solver examples/mnist/lenet_solver.prototxt
# train on GPU 2
caffe train -solver examples/mnist/lenet_solver.prototxt -gpu 2
# train on GPUs 0 & 1 (doubling the batch size)
caffe train -solver examples/mnist/lenet_solver.prototxt -gpu 0,1
# train on all GPUs (multiplying batch size by number of devices)
caffe train -solver examples/mnist/lenet_solver.prototxt -gpu all
# resume training from the half-way point snapshot
caffe train -solver examples/mnist/lenet_solver.prototxt -snapshot examples/mnist/lenet_iter_5000.solverstate
# fine-tune CaffeNet model weights for style recognition
caffe train -solver examples/finetuning_on_flickr_style/solver.prototxt -weights models/bvlc_reference_caffenet/bvlc_reference_caffenet.caffemodel

Testing

# score the learned LeNet model on the validation set as defined in the
# model architeture lenet_train_test.prototxt
caffe test -model examples/mnist/lenet_train_test.prototxt -weights examples/mnist/lenet_iter_10000.caffemodel -gpu 0 -iterations 100

Benchmarking

# (These example calls require you complete the LeNet / MNIST example first.)
# time LeNet training on CPU for 10 iterations
caffe time -model examples/mnist/lenet_train_test.prototxt -iterations 10
# time LeNet training on GPU for the default 50 iterations
caffe time -model examples/mnist/lenet_train_test.prototxt -gpu 0
# time a model architecture with the given weights on the first GPU for 10 iterations
caffe time -model examples/mnist/lenet_train_test.prototxt -weights examples/mnist/lenet_iter_10000.caffemodel -gpu 0 -iterations 10

Diagnostics

# query the first device
caffe device_query -gpu 0

性能

K40（ECC off减少内存占用，boost clock enabled加快始终频率，cuDNN） Traning：19.2s/20iter 5120 imgs Testing：60.7s 50k imgs

http://caffe.berkeleyvision.org/performance_hardware.html

安装

http://caffe.berkeleyvision.org/installation.html

课程

http://vision.stanford.edu/teaching/cs231n/index.html http://neuralnetworksanddeeplearning.com/