plot fscore graph

sum_ind/plot_results.py

+import h5py
+from matplotlib import pyplot as plt
+import argparse
+import os
+import os.path as osp
+import numpy as np
+
+parser = argparse.ArgumentParser()
+parser.add_argument('-p', '--path', type=str,
+                    default='log/non overlapping summe- SumInd (leaky-relu ) and (rep75+div25) layers 2/summe-split4/result.h5',
+                    help="path to h5 file containing summarization results")
+args = parser.parse_args()
+
+parser.add_argument('-d', '--dataset', type=str, default='datasets/eccv16_dataset_summe_google_pool5.h5',
+                    help="path to h5 file containing summarization results")
+args = parser.parse_args()
+
+h5_res = h5py.File(args.path, 'r')
+dataset = h5py.File(args.dataset, 'r')
+keys = h5_res.keys()
+
+for key in keys:
+    score = h5_res[key]['score'][...]
+    machine_summary = h5_res[key]['machine_summary'][...]
+    gtscore = h5_res[key]['gtscore'][...]
+    reference_summaries = np.asarray(dataset[key]['user_summary']).mean(axis=0)
+    fm = h5_res[key]['fm'][()]
+    n = len(reference_summaries)
+    # plot score vs gtscore
+
+    fig, ax = plt.subplots()
+    line1, = ax.plot(range(n), reference_summaries, label='mean reference summaries')
+
+    line2, = ax.plot(range(n), machine_summary, label='generated summary')
+
+    ax.set_xlim(0, n)
+    ax.set_ylim(0, 1.1)
+    ax.set_yticks([0, .2, .4, .6, .8, 1])
+
+    ax.set_title("generated summary for {},  with F-score {:.1%}".format(key, fm))
+    ax.grid()
+    ax.legend()
+
+    fig.savefig(osp.join(osp.dirname(args.path), 'score_' + key + '.png'), bbox_inches='tight')
+    plt.close()
+
+    print("Done video {}. # frames {}.".format(key, len(machine_summary)))
+
+h5_res.close()