time_plot.py

#!/usr/bin/env python
__author__    = 'Danelle Cline'
__copyright__ = '2018'
__license__   = 'GPL v3'
__contact__   = 'dcline at mbari.org'
__doc__ = '''

Combine all GPU/CPU inference times into a single plot
@var __date__: Date of last svn commit
@undocumented: __doc__ parser
@status: production
@license: GPL
'''

import matplotlib
matplotlib.use('Agg')
from pylab import *
import os
import pandas as pd
import glob
import model_metadata as meta
import matplotlib.pyplot as plt
plt.style.use('ggplot')
plt.rcParams['font.family'] = 'serif'
plt.rcParams['font.size'] = 30

if __name__ == '__main__':

    compute = ['GPU'] #['CPU', 'GPU']
    custom_colors = list(['b'])

    for c in compute:
      base_dir = os.path.join(os.getcwd(), 'data')
      #all_files = sorted(glob.iglob(base_dir + '/**/*{0}*.csv'.format(c), recursive=True))
      all_files = sorted(glob.iglob(base_dir + '/**/*.csv', recursive=True))
      df = pd.DataFrame()
      for f in all_files:
        try:
          data = pd.read_csv(f,index_col=None)
          d = os.path.basename(f)
          model_name = d.split('-')[1]
          m = meta.ModelMetadata(model_name)
          if data.empty:
            continue
          if int(m.proposals) > 0:
            model_description = '{0}\n{1} Box Proposals'.format(m.meta_arch, m.proposals)
          elif m.meta_arch == 'SSD':
            model_description = '{0}\n{1} Image Resolution'.format(m.meta_arch, m.image_resolution)
          else:
            model_description = m.meta_arch
          df = df.append({'Model':model_description , '{0} Time'.format(c):int(data.iloc[0]['GPU Time'])}, ignore_index=True)
        except Exception as ex:
          import pdb;pdb.set_trace()
          continue

      with plt.style.context('ggplot'):

        # start a new figure - size is in inches
        fig = plt.figure(figsize=(6, 4), dpi=400)
        ax = plt.subplot()
        df = df.set_index('Model')
        df.plot(kind='bar', sort_columns=True, alpha=0.75, rot=90, legend=False, ax=ax, color=custom_colors)
        ax.set_xlabel("")
        ax.set_ylabel("Milliseconds")
        ax.set_title(label='Average {0} Detection Time'.format(c))
        plt.show()
        fname=fname='{0}_time_plot.png'.format(c)
        fig.savefig(fname=fname, bbox_inches='tight')
        print('Done creating {0}'.format(fname))