cafffe 利用h5py生成多标签h5文件并训练

h5py生成多标签h5文件

import h5py
import numpy as np
def main():
  f = h5py.File('train00.h5', 'w')
  f.create_dataset('data', (1200, 128), dtype='f8')
  f.create_dataset('label', (1200, 4), dtype='i')

  for i in range(1200):
      a = np.empty(128)
      if i % 4 == 0:
          for j in range(128):
              a[j] = j / 128.0;
          l = [1,0,0,0]
      elif i % 4 == 1:
          for j in range(128):
              a[j] = (128 - j) / 128.0;
          l = [1,0,1,0]
      elif i % 4 == 2:
          for j in range(128):
              a[j] = (j % 6) / 128.0;
          l = [0,1,1,0]
      elif i % 4 == 3:
          for j in range(128):
              a[j] = (j % 4) * 4 / 128.0;
          l = [1,0,1,1]
      f['data'][i] = a
      f['label'][i] = l
  f.close()
  with open('train.h5list','w') as f:
      f.write('train00.h5')

  with open('test.h5list','w') as f:
      f.write('train00.h5')

if __name__=="__main__":
  main()

训练数据

import caffe
from caffe import layers as L
from matplotlib import pyplot as plt
import os
import numpy as np

def change_env():
  root = os.path.dirname(__file__)
  os.chdir(root)
  print ("current work root:->",root)

def net(hdf5, batch_size):
  network = caffe.NetSpec()
  network.data,network.label = L.HDF5Data(batch_size=batch_size,source=hdf5,ntop=2)
  network.ip1  = L.InnerProduct(network.data,num_output=50,weight_filler=dict(type="xavier"))
  network.relu1 = L.ReLU(network.ip1,in_place=True)
  network.ip2  = L.InnerProduct(network.relu1,num_output=4,weight_filler=dict(type="xavier"))
  network.loss = L.SigmoidCrossEntropyLoss(network.ip2,network.label)

  return network.to_proto()

def file_write(path=""):
  with open('auto_train01.prototxt', 'w') as f:
    f.write(str(net('train.h5list', 100)))
  with open('auto_test01.prototxt', 'w') as f:
    f.write(str(net('test.h5list', 50)))

def main():
  #change_env()
  file_write()
  solver = caffe.SGDSolver('auto_solver01.prototxt')
  niter = 1001
  test_interval = 10
  train_loss = np.zeros(niter)
  test_acc =  np.zeros(int(np.ceil(niter * 1.0 / test_interval)))
  train_acc =  np.zeros(niter)

  # iteration and save the loss, accuracy
  for it in range(niter):
    solver.step(1)  
    train_loss[it] = solver.net.blobs['loss'].data
    solver.test_nets[0].forward(start='data')

    if it % test_interval == 0:
      print ('Iteration', it, 'testing...')
      correct = 0
      data = solver.test_nets[0].blobs['ip2'].data
      label = solver.test_nets[0].blobs['label'].data
      for test_it in range(100):
        solver.test_nets[0].forward()
        # Positive -> label 1, negative -> label 0
        for i in range(len(data)):
          for j in range(len(data[i])):
            if data[i][j] > 0 and label[i][j] == 1:
              correct += 1
            elif data[i][j] <= 0 and label[i][j] == 0:
              correct += 1
      test_acc[int(it / test_interval)] = correct * 1.0 / (len(data) * len(data[0]) * 100)

  #output graph
  _, ax1 = plt.subplots()
  ax2 = ax1.twinx()
  ax1.plot(np.arange(niter), train_loss)
  ax2.plot(test_interval * np.arange(len(test_acc)), test_acc, 'r')
  ax1.set_xlabel('iteration')
  ax1.set_ylabel('train loss')
  ax2.set_ylabel('test accuracy')
  _.savefig('converge01.png')

if __name__ == '__main__':
  main()