唐宇迪深度学习框架Caffe系列-4

多label的数据

1.jpg 11 78 62 71 84 65 26 51 46 89
2.jpg 13 25 62 53 53 65 26 32 42 13
3.jpg 15 78 66 13 84 65 32 41 51 65

以前是这样的结构

1.jpg 0 或者 2.jpg 1

现在是这样

1.jpg x y  x y  x y  x y  x y  x y

对应我们这里的例子左眼(x,y) 右眼(x,y) 鼻子(x,y) 左嘴角(x,y) 右嘴角(x,y)

HDF5的数据文件制作

import h5py
import os
import cv2
import math
import numpy as np
import random
import re

root_path = "/home/tyd/caffe_case/HDF5/image"

with open("/home/tyd/caffe_case/HDF5/hdf5.txt", 'r') as f:
    lines = f.readlines()

num = len(lines)
random.shuffle(lines)


imgAccu = 0
imgs = np.zeros([num, 3, 224, 224])
labels = np.zeros([num, 10])
for i in range(num):
    line = lines[i]
    segments = re.split('\s+', line)[:-1]
    print segments[0]
    img = cv2.imread(os.path.join(root_path, segments[0])) #取图片路径
    img = cv2.resize(img, (224, 224))
    img = img.transpose(2,0,1) #opencv是H*W*C而Caffe是C*H*W
    imgs[i,:,:,:] = img.astype(np.float32)
    for j in range(10):
        labels[i,j] = float(segments[j+1])*224/256 #由于图像进行了resize，label位置也要发生变换

batchSize = 1 #每一个.h5文件别存放太多图像,分成多个.h5,这里我们每一个存8000个图像
batchNum = int(math.ceil(1.0*num/batchSize))

imgsMean = np.mean(imgs, axis=0)
#imgs = (imgs - imgsMean)/255.0
labelsMean = np.mean(labels, axis=0)
labels = (labels - labelsMean)/10

if os.path.exists('trainlist.txt'):
    os.remove('trainlist.txt')
if os.path.exists('testlist.txt'):
    os.remove('testlist.txt')
comp_kwargs = {'compression': 'gzip', 'compression_opts': 1}
for i in range(batchNum):
    start = i*batchSize
    end = min((i+1)*batchSize, num)
    if i < batchNum-1:
        filename = '/home/tyd/caffe_case/HDF5/h5/train{0}.h5'.format(i)
    else:
        filename = '/home/tyd/caffe_case/HDF5/h5/test{0}.h5'.format(i-batchNum+1)
    print filename
    with h5py.File(filename, 'w') as f:
        f.create_dataset('data', data = np.array((imgs[start:end]-imgsMean)/255.0).astype(np.float32), **comp_kwargs)
        f.create_dataset('label', data = np.array(labels[start:end]).astype(np.float32), **comp_kwargs)

    if i < batchNum-1:
        with open('/home/tyd/caffe_case/HDF5/h5/trainlist.txt', 'a') as f:
            f.write(os.path.join(os.getcwd(), 'train{0}.h5').format(i) + '\n')
    else:
        with open('/home/tyd/caffe_case/HDF5/h5/testlist.txt', 'a') as f:
            f.write(os.path.join(os.getcwd(), 'test{0}.h5').format(i-batchNum+1) + '\n')

imgsMean = np.mean(imgsMean, axis=(1,2))
with open('mean.txt', 'w') as f:
    f.write(str(imgsMean[0]) + '\n' + str(imgsMean[1]) + '\n' + str(imgsMean[2]))

数据预处理可以减去mean值(希望数据以0为中心化)，data和label都可以，但是最后预测的时候要加上mean值 

还写入mean.txt

109.267684418
109.988938935
116.414926127