show_classifier_graphs.py

# -*- coding: utf-8 -*-
"""
Created on Wed Nov  9 17:03:26  2022



@author: Antonio Vispi
"""

import argparse
import os
import json
import matplotlib.pyplot as plt
import numpy as np


def load_json_arr(json_path):
    lines = []
    with open(json_path, 'r') as f:
        for line in f:
            lines.append(json.loads(line))
    return lines

def training_graphs(path_in_json,path_out):
    
    os.makedirs(path_out, exist_ok = True)

    experiment_metrics = load_json_arr(path_in_json)

    Val_Losses=[]
    Train_Losses=[]
    for sample in experiment_metrics[1:len(experiment_metrics)]:
      if sample['mode'] == 'val':
        if 'loss' in sample:
          Val_Losses.append(sample['loss'])
      if sample['mode'] == 'train':
        if 'loss' in sample:
          Train_Losses.append(sample['loss'])

    Val_Accuracy=[]
    Train_Accuracy=[]
    for sample in experiment_metrics[1:len(experiment_metrics)]:
      if sample['mode'] == 'val':
        if 'accuracy_top-1' in sample:
          Val_Accuracy.append(sample['accuracy_top-1'])
      if sample['mode'] == 'train':
        if 'top-1' in sample:
          Train_Accuracy.append(sample['top-1'])
  
    #Definition of the abscissa axes. 
    Epoch_Train = np.linspace(1, len(Val_Losses), num=len(Train_Losses))
    Epoch_Val = np.linspace(1, len(Val_Losses), num=len(Val_Losses))

    fig, axs = plt.subplots(1,2, figsize=(20, 15))
    plt.rcParams['font.size'] = 18

    plt.sca(axs[0])
    plt.yticks(np.arange(0, max(max(Val_Losses),max(Train_Losses)), round(max(max(Val_Losses)/50,max(Train_Losses)/50),2)))
    axs[0].plot(Epoch_Val,Val_Losses,Epoch_Train,Train_Losses)
    axs[0].set_title('Model Losses')
    axs[0].set(xlabel='Epoch')
    axs[0].set(ylabel='Loss')
    axs[0].legend(('Validation', 'Training'), loc='upper right', shadow=True)

    plt.sca(axs[1])
    plt.yticks(np.arange(0, max(max(Val_Accuracy),max(Train_Accuracy)), 5))
    axs[1].plot(Epoch_Val,Val_Accuracy,Epoch_Train,Train_Accuracy)
    axs[1].set_title('Model Accuracy')
    axs[1].set(xlabel='Epoch')
    axs[1].set(ylabel='Accuracy(%)')
    axs[1].legend(('Validation', 'Training'), loc='lower right', shadow=True)
    
    plt.savefig(path_out+'/Graphs.png')
    print('\n')
    print('The graph has been saved in the destination path.')
    

def main():
    parser = argparse.ArgumentParser()
    parser.add_argument('--path_in_json', help='Enter the path of the .json file containing all the annotations of the model training data')
    parser.add_argument('--path_out', help='The path in which the graph summarizing the training data will be saved.')
    
    args = parser.parse_args()
    training_graphs(args.path_in_json,args.path_out)

if __name__ == '__main__':
    main()