File size: 12,416 Bytes
747451d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
# /*---------------------------------------------------------------------------------------------
#  * Copyright (c) 2024 STMicroelectronics.
#  * All rights reserved.
#  * This software is licensed under terms that can be found in the LICENSE file in
#  * the root directory of this software component.
#  * If no LICENSE file comes with this software, it is provided AS-IS.
#  *--------------------------------------------------------------------------------------------*/

import os
import sys
import argparse
from pathlib import Path
import hydra
from hydra.core.hydra_config import HydraConfig
from munch import DefaultMunch
from omegaconf import OmegaConf
from omegaconf import DictConfig
from glob import glob
from tqdm import tqdm
from collections import Counter
from statistics import mean
import matplotlib.pyplot as plt
import numpy as np
import tensorflow as tf


def get_config(config: DictConfig) -> DefaultMunch:
    """
    Converts the configuration data

    Args:
        config (DictConfig): dictionary containing the entire configuration file.

    Returns:
        DefaultMunch: The configuration object.
    """
    config_dict = OmegaConf.to_container(config)
    cfg = DefaultMunch.fromDict(config_dict)
    return cfg


def parse_label_file(txt_file_path : str=None) -> list:
    """
    Provides detections in a list from input text file

    Args:
        txt_file_path (str) : Path of the detection file to analyze

    Returns:
        List : list of detected labels
    """
    labels = []
    if os.path.isfile(txt_file_path):
        with open(txt_file_path, "r") as f:
            data = f.readlines()   
        for line in data:
            if line.rstrip() != "":
                fields = line.split()
                labels.append([float(x) for x in fields])
    return labels


def compute_labels_stats(dataset_path : str=None, 
                         dataset_name : str=None, 
                         histogram_dir: str=None) -> None:
    """
    Provides statistics on the dataset labels

    Args:
        dataset_path (str) : Path of the dataset to analyze
        dataset_name (str) : Name of the dataset used
        histogram_dir (str): location of the histograms storage

    Returns:
        None
    """
    print("\nCalculating groundtruth labels statistics:")
    print("-----------------------------------------")
    print("Dataset root:", dataset_path)

    jpg_file_paths = glob(os.path.join(dataset_path, "*.jpg"))
    if len(jpg_file_paths) == 0:
        raise ValueError(f"Could not find any .jpg file in dataset root directory")

    num_jpg_files = len(jpg_file_paths)
    num_txt_files = 0
    num_empty_txt = 0

    label_sizes = []
    for jpg_path in tqdm(jpg_file_paths):
        txt_path = os.path.join(Path(jpg_path).parent, Path(jpg_path).stem + ".txt")
        if os.path.isfile(txt_path):
            num_txt_files += 1
            labels = parse_label_file(txt_path)
            if not labels:
                num_empty_txt += 1
                label_sizes.append(0)
            else:
                label_sizes.append(len(labels))
#            label_sizes.append(len(labels))

    print("Image files: ", num_jpg_files)
    print("Labels files:", num_txt_files)
    print("Empty labels files:", num_empty_txt)
    print("Labels per image:  min = {}, max = {}, mean = {:.2f}".
                format(min(label_sizes), max(label_sizes), mean(label_sizes)))

    plt.figure(figsize=(8, 8))
    plt.hist(label_sizes, bins=max(label_sizes))
    plot_title = "Number of labels per image"
    if dataset_name:
        plot_title += " in dataset " + dataset_name
    plt.title(plot_title)
    if histogram_dir:
        if not os.path.isdir(histogram_dir):
            os.makedirs(histogram_dir, exist_ok=True)
        plt.savefig(os.path.join(histogram_dir, "labels_stats_" + dataset_name + ".png"))
    plt.show()
    plt.close()


def compute_class_stats(dataset_path : str=None, 
                        dataset_name : str=None, 
                        histogram_dir: str=None) -> None:
    """
    Provides statistics on the dataset classes

    Args:
        dataset_path (str) : Path of the dataset to analyze
        dataset_name (str) : Name of the dataset used
        histogram_dir (str): location of the histograms storage

    Returns:
        None
    """
    print("\nCalculating groundtruth class statistics:")
    print("----------------------------------------")
    print("Dataset root:", dataset_path)
    
    jpg_file_paths = glob(os.path.join(dataset_path, "*.jpg"))
    if len(jpg_file_paths) == 0:
        raise ValueError(f"Could not find any .jpg file in dataset root directory")

    classes = []
    for jpg_path in tqdm(jpg_file_paths):
        txt_path = os.path.join(Path(jpg_path).parent, Path(jpg_path).stem + ".txt")
        if not os.path.isfile(txt_path):
            continue
        labels = parse_label_file(txt_path)
        # Skip .txt files with no objects
        if len(labels) == 0:
            continue
        for i in range(len(labels)):
            id = int(labels[i][0])
            classes.append(id)
            
    classes_dict = Counter(classes)

    class_ids = list(classes_dict.keys())
    class_ids.sort()
    num_classes = max(class_ids) + 1

    print("Number of classes:", num_classes)
    print("Occurences:")
    class_occurences = []
    for id in range(num_classes):
        n = classes_dict[id] if id in classes_dict else 0
        class_occurences.append(n)
        print(f"Class {id}: {n}")

    plt.figure(figsize=(8, 8))
    plot_title = "Class occurences"
    if dataset_name:
        plot_title += " in dataset " + dataset_name
    plt.title(plot_title)
    plt.xticks(class_ids)
    plt.bar(class_ids, class_occurences, width=0.4)
    if histogram_dir:
        if not os.path.isdir(histogram_dir):
            os.makedirs(histogram_dir, exist_ok=True)
        plt.savefig(os.path.join(histogram_dir, "classes_stats_" + dataset_name + ".png"))
    plt.show()
    plt.close()

    
def num_labels_above_cutoff(dataset_path : str=None, 
                            padded_labels_size : int=15) -> float:
    """
    Calculates the percentage of filtered images corresponding to the maximum number of detections 
    kept per image

    Args:
        dataset_path (str): Path of the dataset to analyze
        padded_labels_size (int) : The max number of detection allowed per image 

    Returns:
        float : The corresponding percentage of filtered detections
    """
    print("\nCalculating number of truncated groundtruth labels:")
    print("--------------------------------------------------")
    print("Dataset root:", dataset_path)

    if (padded_labels_size <= 0):
        print("Please make sure that you provided maximum number of detections bigger than 0")
        print("Exiting the script...")
        sys.exit()        

    jpg_file_paths = glob(os.path.join(dataset_path, "*.jpg"))
    if len(jpg_file_paths) == 0:
        raise ValueError(f"Could not find any .jpg file under dataset root {dataset_path}")

    num_examples = 0
    above_cutoff = 0
    txt_file_paths = glob(os.path.join(dataset_path, "*.txt"))
    for path in txt_file_paths:
        num_examples += 1
        labels = parse_label_file(path)
        if len(labels) > padded_labels_size:
            above_cutoff += 1

    cutoff_percentage = 100 * above_cutoff/num_examples
    print("Padded labels size:", padded_labels_size)
    print("Examples with a number of labels greater than padding size: {}/{}  ({:.2f}%)".
          format(above_cutoff, num_examples, cutoff_percentage))
    
    return (cutoff_percentage)


def num_labels_above_percentage(dataset_path : str=None, 
                                target_percentage : float=0.0) -> int:
    """
    Calculates the maximum number of detections in the input images corresponding to the max percentage 
    of the dataset to be filtered by removing images with a lot a detections

    Args:
        dataset_path (str) : Path of the dataset to analyze
        target_percentage (float) : The max percentage of the dataset to be filtered by 
                                    removing images with a lot a detections

    Returns:
        int : The corresponding maximum number of detections per image filtered.
    """
    print("\nCalculating number of truncated groundtruth labels:")
    print("--------------------------------------------------")
    print("Dataset root:", dataset_path)
    if (target_percentage < 0.0) and (target_percentage >= 100.0):
        print("Please make sure that you provided maximum percentage of images to filter between [0.0  100[")
        print("Exiting the script...")
        sys.exit()    

    jpg_file_paths = glob(os.path.join(dataset_path, "*.jpg"))
    if len(jpg_file_paths) == 0:
        raise ValueError(f"Could not find any .jpg file under dataset root {dataset_path}")

    num_examples = 0
    label_sizes = []
    txt_file_paths = glob(os.path.join(dataset_path, "*.txt"))
    for path in txt_file_paths:
        num_examples += 1
        labels = parse_label_file(path)
        if not labels:
            label_sizes.append(0)
        else:
            label_sizes.append(len(labels))

    padded_labels_size = max(label_sizes)
    above_cutoff_final = 0
    while padded_labels_size > 0:
        above_cutoff = 0
        for path in txt_file_paths:
            labels = parse_label_file(path)
            if len(labels) > padded_labels_size:
                above_cutoff += 1
        current_percentage = 100 * above_cutoff/num_examples
        if (current_percentage <= target_percentage):
            above_cutoff_final = above_cutoff
            padded_labels_size -= 1
        else:
            above_cutoff = above_cutoff_final
            break

    print("Padded labels size:", padded_labels_size+1)
    print("Examples with a number of labels greater than padding size: {}/{}  ({:.2f}%)".
          format(above_cutoff, num_examples, 100 * above_cutoff/num_examples))

    return (padded_labels_size+1)


@hydra.main(version_base=None, config_path="", config_name="dataset_config")
def main(configs: DictConfig) -> None:
    """
    Main entry point of the script.
 
    Args:
        cfg: Configuration dictionary.
 
    Returns:
        None
    """
    cfg = get_config(configs)
    cfg.output_dir = HydraConfig.get().run.dir
    if not "dataset_path" in cfg.dataset:
        print("Please make sure that you provided a dataset path")
        print("Exiting the script...")
        sys.exit()
 
    # Set default values of missing optional attributes
    if not cfg.dataset.dataset_name:
        cfg.dataset.dataset_nameme = "<unnamed>"

    # Compute classes statistics
    compute_class_stats(dataset_path = cfg.dataset.dataset_path, 
                        dataset_name = cfg.dataset.dataset_name, 
                        histogram_dir = "histograms")

    # Compute labels statistics
    compute_labels_stats(dataset_path = cfg.dataset.dataset_path, 
                         dataset_name = cfg.dataset.dataset_name, 
                         histogram_dir = "histograms")

    # Provide statistics using max number of detections per image
    if "max_detections" in cfg.settings and (cfg.settings.max_detections != None) :
        max_detections_percentage_filtered = num_labels_above_cutoff(dataset_path = cfg.dataset.dataset_path, 
                                                padded_labels_size = cfg.settings.max_detections)

    # Provide statistics using max number of detections per image
    if "max_percentage_filtered" in cfg.settings and (cfg.settings.max_percentage_filtered != None) :
        max_detections_allowed = num_labels_above_percentage(dataset_path = cfg.dataset.dataset_path,
                                                             target_percentage = cfg.settings.max_percentage_filtered)


if __name__ == "__main__":
    parser = argparse.ArgumentParser()
    parser.add_argument('--config-path', type=str, default='', help='Path to folder containing configuration file')
    parser.add_argument('--config-name', type=str, default='user_config', help='name of the configuration file')
    
    # Add arguments to the parser
    parser.add_argument('params', nargs='*',
                        help='List of parameters to over-ride in config.yaml')
    args = parser.parse_args()

    # Call the main function
    main()