draw_prec-recall

import sys
import argparse
import matplotlib.pyplot as plt
from pylab import *
import numpy as np

def compare(line1,line2):
    print ("line1 = ",line1)
    print ("line2 = ",line2)

    iter_num1  = int(line1.split()[0])
    iter_num2  = int(line2.split()[0])
    if iter_num1<iter_num2:
        return -1
    elif iter_num1>iter_num2:
        return 1
    else :
        return 0

def main(argv):
    parser = argparse.ArgumentParser()
    parser.add_argument('input_file',default = "C:\\darknet_fire_detection\\build\\darknet\\x64\\log\\precision_recall5.txt", help="file to read precision and recall")

    args = parser.parse_args()
    args.output_file = args.input_file.replace(".txt",".png")

    print ("input file you provided is {}".format(args.input_file))
    print ("output file you provided is {}".format(args.output_file))

    f = open(args.input_file)

    lines = [line.rstrip("\n") for line in f.readlines()]
    print (lines)
    f.close()


    iters = []
    fire_precisions = []
    fire_recalls = []

    smoke_precisions = []
    smoke_recalls = []

    for line in lines:
        cols = line.split()

        iters.append(float(cols[0][:-1]))
        fire_precisions.append(float(cols[1]))

        if 'nan' in cols[2]:
            cols[2]='0'
        fire_recalls.append(float(cols[2]))

        smoke_precisions.append(float(cols[3]))

        if 'nan' in cols[4]:
            cols[4]='0'
        smoke_recalls.append(float(cols[4]))

    fire_F1s = 2*np.array(fire_precisions)*np.array(fire_recalls)/(np.array(fire_precisions)+np.array(fire_recalls))
    smoke_F1s = 2*np.array(smoke_precisions)*np.array(smoke_recalls)/(np.array(smoke_precisions)+np.array(smoke_recalls))

    fig= plt.figure()
    fig.set_size_inches(8,6)
    ax = fig.add_subplot(221)
    plt.plot(iters,fire_precisions,label = "fire precision")
    plt.plot(iters,fire_recalls, label = "fire recall")
    plt.plot(iters,fire_F1s, label = "fire F1")
    plt.legend()
    ax.set_yticks(np.linspace(0,1,11))
    plt.grid()
    plt.ylabel("precision/recall value. Ideally should be 1")

    plt.legend()
    ax.set_yticks(np.linspace(0,1,11))

    ax = fig.add_subplot(222)
    plt.plot(iters,smoke_precisions,label = "smoke precision")
    plt.plot(iters,smoke_recalls, label = "smoke recall")
    plt.plot(iters,smoke_F1s, label = "smoke F1")
    plt.legend()
    ax.set_yticks(np.linspace(0,1,11))
    plt.grid()



    ax = fig.add_subplot(223)

    avg_F1s = 2*(np.array(fire_F1s)*np.array(smoke_F1s))/(np.array(fire_F1s)+np.array(smoke_F1s))

    plt.plot(iters,avg_F1s, color='g', label = "F1 of F1s")
    ax.set_yticks(np.linspace(0,1,11))
    plt.grid()
    plt.xlabel("number of iterations in K(1.0 means 1000)")
    plt.legend()

    savefig(args.output_file,dpi=200)
    plt.show()

if __name__ == "__main__":
    main(sys.argv)