Caffe均值文件binaryproto转npy

原文：

http://blog.csdn.net/langb2014/article/details/52995968

http://blog.csdn.net/may0324/article/details/52316967

之前做过点项目，整理一下小细节。

mean.binaryproto转mean.npy

首先是生成binaryproto的sh文件

[cpp] view plain copy

1. #!/usr/bin/env sh  
2. # Compute the mean image from the imagenet training lmdb  
3. # N.B. this is available in data/ilsvrc12  
4. EXAMPLE=.  
5. DATA=/data/model_train/trainSet  
6. TOOLS=/data/cuda+digits3.0+cudnnv5+opencv/caffe-master/build/tools  
7.   
8. $TOOLS/compute_image_mean $EXAMPLE/train_lmdb \  
9.   $EXAMPLE/train_lmdb/trainset_mean.binaryproto  
10.   
11. $TOOLS/compute_image_mean $EXAMPLE/test_lmdb \  
12.   $EXAMPLE/test_lmdb/testset_mean.binaryproto  
13.   
14. echo "Done."  

然后调用python脚本转换为npy数据

[python] view plain copy

1. #!/usr/bin/env python  
2. #coding=gbk  
3. import numpy as np  
4. import os  
5. import sys  
6. import argparse  
7. import glob  
8. import time,datetime  
9. import scipy.io as scio  
10.   
11. caffe_root = '/data/cuda+digits3.0+cudnnv5+opencv/caffe-master/' #caffe-master            
12. import sys  
13. pycaffe_dir=sys.path.insert(0, caffe_root + 'python')  
14. import caffe  
15. caffe.set_mode_gpu();  
16.   
17. if len(sys.argv)!=3:  
18.     print "Usage: python convert_mean.py mean.binaryproto mean.npy"  
19.     sys.exit()  
20.   
21. blob = caffe.proto.caffe_pb2.BlobProto()  
22. bin_mean = open( sys.argv[1] , 'rb' ).read()  
23. blob.ParseFromString(bin_mean)  
24. arr = np.array( caffe.io.blobproto_to_array(blob) )  
25. npy_mean = arr[0]  
26. np.save( sys.argv[2] , npy_mean )  

在终端使用方法

[python] view plain copy

1. python convert_mean.py mean.binaryproto mean.npy 
   

其他方法

binaryproto -> npy

import numpy as np
import caffe
import sys

blob = caffe.proto.caffe_pb2.BlobProto()
data = open( 'mean.binaryproto' , 'rb' ).read()
blob.ParseFromString(data)
arr = np.array( caffe.io.blobproto_to_array(blob) )
out = arr[0]
np.save( 'mean.npy' , out )

npy -> binaryproto
data_mean is 1*H*W

import numpy as np
import caffe
import sys

blob = caffe.proto.caffe_pb2.BlobProto()
with open('mean.npy','rb') as f:
     data_mean = numpy.load(f)
blob.channels=1
blob.height = data_mean.shape[0]
blob.width = data_mean.shape[1]
blob.data.extend(data_mean.astype(float).flat)
binaryproto_file = open('mean.binaryproto', 'wb' ) 
binaryproto_file.write(blob.SerializeToString())
binaryproto_file.close()