minist_plot

import numpy as np
import re
import os
import matplotlib.pyplot as plt
from pandas import *


laber = '60000/60000'

def save_data(file_dir):
    data_train_acc = []
    data_test_acc = []
    fr = open(file_dir)
    for line in fr.readlines():
        if laber in line:
            match = re.search(r'acc: (\S+) - val', line)
            data_train_acc.append(float(match.group(1)))
            match = re.search(r' - val_acc: (\S+)', line)
            data_test_acc.append(float(match.group(1)))
    fr.close()
    return data_train_acc , data_test_acc

data_train_acc_to = []
data_test_acc_to = []
file_name = []
dir = 'D:\mobiledata\minist_cnntest\minist_result'
for file in os.listdir(dir):
    if '.log' in str(file):
        file_name.append(file.split('.')[0])
        file_path = os.path.join(dir,file)

        data_acc = save_data(file_path)
        data_train_acc_to.append(data_acc[0])
        data_test_acc_to.append(data_acc[1])
        x = [a for a in range(1,len(data_acc[0])+1)]
        # plt.figure()
        # plt.plot(x, data_acc[0], color='blue', label='train_acc')
        # plt.plot(x, data_acc[1], color='red', label='test_acc')
        # plt.legend(loc='lower right')
        # plt.xlabel('number_epoch')
        # plt.ylabel('accuracy')
        # file_before = file.split('.')[0]
        # print file_before
        # plt.title(file_before)
        # plt.xlim(1,80)
        # plt.ylim(0.98,1.0)
        # plt.show()
        # plt.savefig(file_before+'.jpg')
# print np.array(data_test_acc_to).T
# print np.array(data_train_acc_to)
df=DataFrame(np.array(data_train_acc_to).T,columns=file_name,index=np.arange(1,81,1))
# plt.show(df.plot())
# print df
# df.to_csv('data2.csv')
df.plot()
plt.xlabel('number_epoch')
plt.ylabel('accuracy')
plt.ylim(0.988,1.002)
plt.title('model train accuracy')
plt.show()