a script to plot train and test curves altogether.

plot_train_test.py

# In the name of GOD the most compassionate the most merciful
# Originally developed by Yasse Souri
# Just added the search for current directory so that users dont have to use command prompts anymore!
# and also shows the top 4 accuracies achieved so far, and displaying the highest in the plot title 
# Added train/test plot!
# Coded By: Seyyed Hossein Hasan Pour ([email protected])
# -------How to Use ---------------
# 0.First of all make sure your test layers name is 'accuracy' and the layer which reports your training accuracy is
# 'accuracy_training'. then you are good to go. if you are using different names for your test/train layers, make sure to edit
# the regex yourself. I might remove this limitation later, but at the moment this is a hacky edition to get my job done!
# so the code isnot polished! but is straight forward!
# 1.Just place your caffe's traning/test log file (with .log extension) next to this script
# and then run the script.If you have multiple logs placed next to the script, it will plot all of them
# you may also copy this script to your working directory, where you generate/keep your train/test logs
# and easily execute the script and see the curve plotted. 
# this script is standalone.
# 2. you can use command line arguments as well, just feed the script with different log files separated by space
# and you are good to go.
#----------------------------------
import numpy as np
import re
import click
import glob, os
from matplotlib import pylab as plt
import operator
import ntpath
@click.command()
@click.argument('files', nargs=-1, type=click.Path(exists=True))
def main(files):
    plt.style.use('ggplot')
    fig, ax1 = plt.subplots()
    ax2 = ax1.twinx()
    ax1.set_xlabel('iteration')
    ax1.set_ylabel('loss')
    ax2.set_ylabel('accuracy %')
    if not files:
        print 'no args found'
        print '\n\rloading all files with .log extension from current directory'
        os.chdir(".")
        files = glob.glob("*.log")

    for i, log_file in enumerate(files):
        loss_iterations, losses, accuracy_iterations, accuracies, accuracies_iteration_checkpoints_ind, fileName = parse_log(log_file)
        disp_results(fig, ax1, ax2, loss_iterations, losses, accuracy_iterations, accuracies, accuracies_iteration_checkpoints_ind, fileName, color_ind=i)
		
        loss_iterations, losses, accuracy_iterations, accuracies, accuracies_iteration_checkpoints_ind, fileName = parse_log2(log_file)
        disp_results(fig, ax1, ax2, loss_iterations, losses, accuracy_iterations, accuracies, accuracies_iteration_checkpoints_ind, fileName, color_ind=i+1)
		
    plt.show()
	

def parse_log2(log_file):
    with open(log_file, 'r') as log_file2:
        log = log_file2.read()

    #loss_pattern = r"Iteration (?P<iter_num>\d+), loss = (?P<loss_val>[+-]?(\d+(\.\d*)?|\.\d+)([eE][+-]?\d+)?)"
    #loss_pattern2 = r"Batch (?P<iter_num>\d+), accuracy = (?P<accuracy>[+-]?(\d+(\.\d*)?|\.\d+)([eE][+-]?\d+)?)(\n.* = (?P<loss>[+-]?(\d+(\.\d*)?|\.\d+)([eE][+-]?\d+)?))";
    losses = []
    loss_iterations = []
    loss_accuracy_pattern = r"Iteration (?P<iter_num>\d+), loss = ((\d*.\d*)+)\n.*( *)Train net output #0: accuracy_training = ((\d*.\d*)+)"
    accuracies = []
    accuracy_iterations = []
    accuracies_iteration_checkpoints_ind = []
     
    fileName= 'train_'+os.path.basename(log_file)
    #print re.search(loss_pattern2,log) 	
    # if re.search(loss_pattern2,log) != None:
        # for r in re.findall(loss_pattern2, log):
            # loss_iterations.append(int(r[0]))
            ##print '\n'
            ##print (r)	
            # losses.append(float(r[6]))
            # iteration = int(r[0])
            # accuracy = float(r[1]) * 100

            # if iteration % 10000 == 0 and iteration > 0:
                # accuracies_iteration_checkpoints_ind.append(len(accuracy_iterations))

            # accuracy_iterations.append(iteration)
            # accuracies.append(accuracy)
		
	
    if re.search(loss_accuracy_pattern,log) != None:
        for r in re.findall(loss_accuracy_pattern, log):
            #print '\n'
            #print (r)	
            iteration = int(r[0])
            loss_iterations.append(iteration)
            losses.append(float(r[1]))
			
            accuracy = float(r[4]) * 100
			
            if iteration % 10000 == 0 and iteration > 0:
                accuracies_iteration_checkpoints_ind.append(len(accuracy_iterations))

            accuracy_iterations.append(iteration)
            accuracies.append(accuracy)
			
        # for r in re.findall(accuracy_pattern, log):
            # iteration = int(r[0])
            # accuracy = float(r[1]) * 100

            # if iteration % 10000 == 0 and iteration > 0:
                # accuracies_iteration_checkpoints_ind.append(len(accuracy_iterations))

            # accuracy_iterations.append(iteration)
            # accuracies.append(accuracy)
        
    loss_iterations = np.array(loss_iterations)
    losses = np.array(losses)

    accuracy_iterations = np.array(accuracy_iterations)
    accuracies = np.array(accuracies)
	
    return loss_iterations, losses, accuracy_iterations, accuracies, accuracies_iteration_checkpoints_ind, fileName

def parse_log(log_file):
    with open(log_file, 'r') as log_file2:
        log = log_file2.read()

    #loss_pattern = r"Iteration (?P<iter_num>\d+), loss = (?P<loss_val>[+-]?(\d+(\.\d*)?|\.\d+)([eE][+-]?\d+)?)"
    #loss_pattern2 = r"Batch (?P<iter_num>\d+), accuracy = (?P<accuracy>[+-]?(\d+(\.\d*)?|\.\d+)([eE][+-]?\d+)?)(\n.* = (?P<loss>[+-]?(\d+(\.\d*)?|\.\d+)([eE][+-]?\d+)?))";
    losses = []
    loss_iterations = []
    loss_accuracy_pattern = r"Iteration (?P<iter_num>\d+), Testing net \(#0\)((.*?)(\n)*)* accuracy = ((\d*.\d*)+)((.*?)(\n)*)*(.?)*loss = ((\d*.\d*)+) \("
    accuracies = []
    accuracy_iterations = []
    accuracies_iteration_checkpoints_ind = []
     
    fileName= 'test_'+os.path.basename(log_file)
    # print re.search(loss_pattern2,log) 	
    # if re.search(loss_pattern2,log) != None:
        # for r in re.findall(loss_pattern2, log):
            # loss_iterations.append(int(r[0]))
            # #print '\n'
            # #print (r)	
            # losses.append(float(r[6]))
            # iteration = int(r[0])
            # accuracy = float(r[1]) * 100

            # if iteration % 10000 == 0 and iteration > 0:
                # accuracies_iteration_checkpoints_ind.append(len(accuracy_iterations))

            # accuracy_iterations.append(iteration)
            # accuracies.append(accuracy)
		
	
    if re.search(loss_accuracy_pattern,log) != None:
        for r in re.findall(loss_accuracy_pattern, log):
            iteration = int(r[0])
            loss_iterations.append(iteration)
            losses.append(float(r[10]))
            #print '\n'
            #print (r)	
            
            accuracy = float(r[4]) * 100

            if iteration % 10000 == 0 and iteration > 0:
                accuracies_iteration_checkpoints_ind.append(len(accuracy_iterations))

            accuracy_iterations.append(iteration)
            accuracies.append(accuracy)
			
        # for r in re.findall(accuracy_pattern, log):
            # #print '\n'
            # #print (r)	
            # iteration = int(r[0])
            # accuracy = float(r[5]) * 100

            # if iteration % 10000 == 0 and iteration > 0:
                # accuracies_iteration_checkpoints_ind.append(len(accuracy_iterations))

            # accuracy_iterations.append(iteration)
            # accuracies.append(accuracy)
        
    loss_iterations = np.array(loss_iterations)
    losses = np.array(losses)

    accuracy_iterations = np.array(accuracy_iterations)
    accuracies = np.array(accuracies)
	
    return loss_iterations, losses, accuracy_iterations, accuracies, accuracies_iteration_checkpoints_ind, fileName


def disp_results(fig, ax1, ax2, loss_iterations, losses, accuracy_iterations, accuracies, accuracies_iteration_checkpoints_ind, fileName, color_ind=0):
    modula = len(plt.rcParams['axes.color_cycle'])
    acrIterations =[]
    top_acrs={}
    if accuracies.size:
        if 	accuracies.size>4:
		    top_n = 4
        else:
            top_n = accuracies.size -1		
        temp = np.argpartition(-accuracies, top_n)
        result_indexces = temp[:top_n]
        temp = np.partition(-accuracies, top_n)
        result = -temp[:top_n]
        for acr in result_indexces:
            acrIterations.append(accuracy_iterations[acr])
            top_acrs[str(accuracy_iterations[acr])]=str(accuracies[acr])

        sorted_top4 = sorted(top_acrs.items(), key=operator.itemgetter(1))
        maxAcc = np.amax(accuracies, axis=0)
        iterIndx = np.argmax(accuracies)
        maxAccIter = accuracy_iterations[iterIndx]
        maxIter =   accuracy_iterations[-1]
        consoleInfo = format('\n[%s]:maximum accuracy [from 0 to %s ] = [Iteration %s]: %s ' %(fileName,maxIter,maxAccIter ,maxAcc))
        plotTitle = format('max accuracy(%s) [Iteration %s]: %s ' % (fileName,maxAccIter, maxAcc))
        print (consoleInfo)
        #print (str(result))
        #print(acrIterations)
       # print 'Top 4 accuracies:'		
        print ('Top 4 accuracies:'+str(sorted_top4))		
        plt.title(plotTitle)
    ax1.plot(loss_iterations, losses, color=plt.rcParams['axes.color_cycle'][(color_ind * 2 + 0) % modula])
    ax2.plot(accuracy_iterations, accuracies, plt.rcParams['axes.color_cycle'][(color_ind * 2 + 1) % modula], label=str(fileName))
    ax2.plot(accuracy_iterations[accuracies_iteration_checkpoints_ind], accuracies[accuracies_iteration_checkpoints_ind], 'o', color=plt.rcParams['axes.color_cycle'][(color_ind * 2 + 1) % modula])
    plt.legend(loc='lower right') 


if __name__ == '__main__':
    main()