Teach yourself to train pytorch model to Caffe (III)

This section uses caffe To reason out pytorh The same result

Reference resources ： Use Caffe Of Python Interface for reasoning - Simple books (jianshu.com)

#coding = utf-8
import numpy as np
import sys, os


sys.path.insert(0, caffe_root + 'python')

import caffe

os.chdir(caffe_root)

#caffe.set_device(0)
#caffe.set_mode_gpu()
#### The following three lines are modified according to their own path 
net_file = './dock_googlenet_dog.prototxt'
caffe_model = './dock_googlenet_dog.caffemodel'
imagenet_labels_filename = './class_names.txt'


# load model
net = caffe.Net(net_file, caffe_model, caffe.TEST)
#resize
transformer = caffe.io.Transformer({'data': net.blobs['data'].data.shape})
# Get the number of channels to the front   Turn into 【1,3 224,224】
transformer.set_transpose('data', (2, 0, 1))
# #  Call the mean file 
# # transformer.set_mean('data', np.load(mean_file).mean(1).mean(1))
#python Store pictures as [0, 1], and caffe Store pictures as [0, 255], So we need a transformation 
# transformer.set_raw_scale('data', 255)
#caffe The picture is RGB Format , And network format BGR, To transform 
# transformer.set_channel_swap('data', (2, 1, 0))
### Change according to your own path 
im = caffe.io.load_image('./cat.2.jpg')
net.blobs['data'].data[...] = transformer.preprocess('data',im)
# net.blobs['data'].data[...] = 

out = net.forward()

print (net.blobs['data'].data.shape)
print (net.blobs['data'].data.dtype)
print (net.blobs[net.blobs.keys()[-1]].data.shape)
print (net.blobs[net.blobs.keys()[-1]].data.dtype)


labels = np.loadtxt(imagenet_labels_filename, str, delimiter='\t')

top_k = net.blobs[net.blobs.keys()[-1]].data[0].flatten().argsort()[-1:-6:-1]
for i in np.arange(top_k.size):
    print (top_k[i], labels[top_k[i]])
# Print the final score 
print (net.blobs[net.blobs.keys()[-1]].data[0])

 ./class_names.txt Namely

0 dog
1 cat

Note that my training is not used normlize,

  And my format is RGB Of , So I have a few useless preprocessing

caffe Of caffe.io.load_image('./cat.2.jpg') yes RGB Of

Mine is also. RBG Of

  If you use brother's docker My code will report three errors when the environment runs

###got an unexpected keyword argument ‘as_grey‘
https://blog.csdn.net/hjxu2016/article/details/113541609
###TypeError: 'float' object cannot be interpreted as an integer
https://blog.csdn.net/qq_28634403/article/details/81224291
###dict_keys‘ object is not subscriptable
https://blog.csdn.net/le___le/article/details/103617431

  I use my own configuration caffe, my caffe No, ceil mode Parameters , use round_mode Instead of the , If ceil mode by true be round_mode by 0, Otherwise, vice versa . Accuracy basically does not affect Pytorch maxpool Of ceil_mode And caffe in maxpool Comparison _zl3090 The blog of -CSDN Blog

How to match yourself caffe Look at my other blog ：

finally ubuntu Finished compiling CPU edition caffe 了 _FeboReigns The blog of -CSDN Blog