Skip to content

Dataset utilities

Kano provides some classes to visualize and manipulate YOLO-formatted datasets.

Note

A valid YOLO-formatted dataset will have a folder tree like this 

    dataset_name:
    ├─ train
    |  ├─ images
    |  |  ├─ image1.jpg
    |  |  └─ ...
    |  └─ labels
    |     ├─ image1.txt
    |     └─ ...
    ├─ valid (optional)
    ├─ test (optional)
    └─ data.yaml (must include a list of classes names inside "names" field)

  • YoloImage: visualize, copy a Yolo-formmated image.
  • YoloDataset: visualize, merge, split Yolo-formatted datasets.

kano.dataset_utils.YoloImage

Represents an image annotated in YOLO format, which includes bounding boxes or skeletons.

Attributes:

Name Type Description
image ndarray

The original image.
image_path str

Path to the image file.
label_path str

Path to the label file corresponding to the image.
task str

Task type, either "detect" or "pose".
labels list

List of dictionaries containing label information.
labels_dict dict

Dictionary mapping class IDs to class names.
Source code in kano\dataset_utils.py 
 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
class YoloImage:
    """
    Represents an image annotated in YOLO format, which includes bounding boxes or skeletons.

    Attributes:
        image (numpy.ndarray): The original image.
        image_path (str): Path to the image file.
        label_path (str): Path to the label file corresponding to the image.
        task (str): Task type, either "detect" or "pose".
        labels (list): List of dictionaries containing label information.
        labels_dict (dict): Dictionary mapping class IDs to class names.
    """

    def __init__(self, image_path, labels_dict=None, task="detect"):
        """
        Initialize a YoloImage object.

        Args:
            image_path (str): Path to the image file.
            labels_dict (dict): Dictionary mapping class IDs to class names.
            task (str): Task type. Possible values: "detect", "pose".
        """
        self.image = cv2.imread(image_path)
        self.image_path = image_path
        if task not in TASKS:
            raise ValueError("Unexpected task. Please provide one of:", TASKS)
        self.task = task
        self.labels = self.get_labels()
        self.labels_dict = labels_dict

    def get_label_path(self, image_path):
        """
        Get the path to the label file corresponding to the given image.

        Args:
            image_path (str): Path to the image file.

        Returns:
            label_path (str): Path to the label file.
        """
        image_path = Path(image_path)
        images_folder_path = image_path.parent
        dataset_path = images_folder_path.parent
        labels_folder_path = dataset_path / "labels"
        label_filename = image_path.with_suffix(".txt").name
        label_path = labels_folder_path / label_filename

        return str(label_path)

    def get_labels(self):
        """
        Parse the label file and extract label information.
        Each label can contain:
            - class: class of the object
            - s_xywh: scaled xywh box
            - xyxy: xyxy box
            - keypoints: list of dict(xy, state) for pose estimation tasks

        Returns:
            labels (list): List of dictionaries, each containing label information.
        """
        self.label_path = self.get_label_path(self.image_path)
        labels = list()
        image_height, image_width = self.image.shape[:2]
        with open(self.label_path, "r") as file:
            for line in file:
                line = line.strip().split()
                label = {
                    "class": int(line[0]),
                    "s_xywh": np.array([float(x) for x in line[1:5]]),
                }
                xywh = label["s_xywh"].copy() * np.array(
                    [image_width, image_height, image_width, image_height]
                )
                label["xyxy"] = xywh2xyxy(xywh)

                if self.task == "pose":
                    keypoints = list()
                    for start_id in range(5, len(line), 3):
                        # "state" in [0, 1, 2] with:
                        # - 0: deleted
                        # - 1: occluded
                        # - 2: visible
                        x = int(float(line[start_id]) * image_width)
                        y = int(float(line[start_id + 1]) * image_height)
                        state = int(float(line[start_id + 2]))
                        keypoints.append(
                            {
                                "xy": (x, y),
                                "state": state,
                            }
                        )
                    label["keypoints"] = keypoints

                labels.append(label)
        return labels

    def show_image(self, figsize=(10, 10)):
        """
        Display the original image.

        Args:
            figsize (tuple(int, int)): Size of the figure (width, height) in inches.
        """
        show_image(self.image, figsize)

    def get_annotated_image(self):
        """
        Get the annotated image with bounding boxes or skeletons drawn.

        Returns:
            annotated_image (np.ndarray): Annotated image.
        """
        annotated_image = self.image.copy()
        for label in self.labels:

            if self.task == "pose":
                annotated_image = draw_skeleton(
                    annotated_image, label["keypoints"]
                )

            cls = label["class"]
            if self.labels_dict is not None:
                cls = self.labels_dict[cls]
            bbox = label["s_xywh"]

            annotated_image = draw_bbox(
                annotated_image, bbox, "s_xywh", (0, 255, 0), str(cls)
            )

        return annotated_image

    def show_annotated_image(self, figsize=(10, 10)):
        """
        Display the annotated image with bounding boxes or skeletons drawn.

        Returns:
            figsize (tuple(int, int)): Size of the figure (width, height) in inches.
        """
        annotated_image = self.get_annotated_image()
        show_image(annotated_image, figsize)

    def copy_to(
        self,
        target_folder_path,
        prefix="",
        reindex_dict=None,
        target_stem=None,
    ):
        """
        Copy the image and its label file to the specified target folder.

        Returns:
            target_folder_path (str): Path to the target folder.
            prefix (str): Prefix to be added to the filenames
            reindex_dict (dict): Dictionary mapping original class IDs to new class IDs.
            target_stem (str): Name stem for the copied files
        """
        source_image_path = Path(self.image_path)
        source_label_path = Path(self.label_path)
        target_folder_path = Path(target_folder_path)

        if target_stem is None:
            target_image_path = (
                target_folder_path
                / "images"
                / (prefix + source_image_path.name)
            )
            target_label_path = (
                target_folder_path
                / "labels"
                / (prefix + source_label_path.name)
            )
        else:
            target_image_path = (
                target_folder_path / "images" / (target_stem + ".jpg")
            )
            target_label_path = (
                target_folder_path / "labels" / (target_stem + ".txt")
            )

        shutil.copyfile(self.image_path, str(target_image_path))
        shutil.copyfile(self.label_path, str(target_label_path))

        if reindex_dict:
            with open(str(source_label_path), "r") as f:
                lines = f.readlines()

            new_lines = list()
            for line in lines:
                class_id = int(line.split()[0])
                if class_id in reindex_dict:
                    new_class_id = reindex_dict[class_id]
                    if new_class_id is not None:
                        new_line = (
                            f"{new_class_id} {' '.join(line.split()[1:])}\n"
                        )
                        new_lines.append(new_line)

            with open(str(target_label_path), "w") as f:
                f.writelines(new_lines)

__init__(image_path, labels_dict=None, task='detect')

Initialize a YoloImage object.

Parameters:

Name Type Description Default
image_path str

Path to the image file.

 required
labels_dict dict

Dictionary mapping class IDs to class names.

 None
task str

Task type. Possible values: "detect", "pose".

 'detect'
Source code in kano\dataset_utils.py 
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
def __init__(self, image_path, labels_dict=None, task="detect"):
    """
    Initialize a YoloImage object.

    Args:
        image_path (str): Path to the image file.
        labels_dict (dict): Dictionary mapping class IDs to class names.
        task (str): Task type. Possible values: "detect", "pose".
    """
    self.image = cv2.imread(image_path)
    self.image_path = image_path
    if task not in TASKS:
        raise ValueError("Unexpected task. Please provide one of:", TASKS)
    self.task = task
    self.labels = self.get_labels()
    self.labels_dict = labels_dict

copy_to(target_folder_path, prefix='', reindex_dict=None, target_stem=None)

Copy the image and its label file to the specified target folder.

Returns:

Name Type Description
target_folder_path str

Path to the target folder.
prefix str

Prefix to be added to the filenames
reindex_dict dict

Dictionary mapping original class IDs to new class IDs.
target_stem str

Name stem for the copied files
Source code in kano\dataset_utils.py 
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
def copy_to(
    self,
    target_folder_path,
    prefix="",
    reindex_dict=None,
    target_stem=None,
):
    """
    Copy the image and its label file to the specified target folder.

    Returns:
        target_folder_path (str): Path to the target folder.
        prefix (str): Prefix to be added to the filenames
        reindex_dict (dict): Dictionary mapping original class IDs to new class IDs.
        target_stem (str): Name stem for the copied files
    """
    source_image_path = Path(self.image_path)
    source_label_path = Path(self.label_path)
    target_folder_path = Path(target_folder_path)

    if target_stem is None:
        target_image_path = (
            target_folder_path
            / "images"
            / (prefix + source_image_path.name)
        )
        target_label_path = (
            target_folder_path
            / "labels"
            / (prefix + source_label_path.name)
        )
    else:
        target_image_path = (
            target_folder_path / "images" / (target_stem + ".jpg")
        )
        target_label_path = (
            target_folder_path / "labels" / (target_stem + ".txt")
        )

    shutil.copyfile(self.image_path, str(target_image_path))
    shutil.copyfile(self.label_path, str(target_label_path))

    if reindex_dict:
        with open(str(source_label_path), "r") as f:
            lines = f.readlines()

        new_lines = list()
        for line in lines:
            class_id = int(line.split()[0])
            if class_id in reindex_dict:
                new_class_id = reindex_dict[class_id]
                if new_class_id is not None:
                    new_line = (
                        f"{new_class_id} {' '.join(line.split()[1:])}\n"
                    )
                    new_lines.append(new_line)

        with open(str(target_label_path), "w") as f:
            f.writelines(new_lines)

get_annotated_image()

Get the annotated image with bounding boxes or skeletons drawn.

Returns:

Name Type Description
annotated_image ndarray

Annotated image.
Source code in kano\dataset_utils.py 
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
def get_annotated_image(self):
    """
    Get the annotated image with bounding boxes or skeletons drawn.

    Returns:
        annotated_image (np.ndarray): Annotated image.
    """
    annotated_image = self.image.copy()
    for label in self.labels:

        if self.task == "pose":
            annotated_image = draw_skeleton(
                annotated_image, label["keypoints"]
            )

        cls = label["class"]
        if self.labels_dict is not None:
            cls = self.labels_dict[cls]
        bbox = label["s_xywh"]

        annotated_image = draw_bbox(
            annotated_image, bbox, "s_xywh", (0, 255, 0), str(cls)
        )

    return annotated_image

get_label_path(image_path)

Get the path to the label file corresponding to the given image.

Parameters:

Name Type Description Default
image_path str

Path to the image file.

 required

Returns:

Name Type Description
label_path str

Path to the label file.
Source code in kano\dataset_utils.py 
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
def get_label_path(self, image_path):
    """
    Get the path to the label file corresponding to the given image.

    Args:
        image_path (str): Path to the image file.

    Returns:
        label_path (str): Path to the label file.
    """
    image_path = Path(image_path)
    images_folder_path = image_path.parent
    dataset_path = images_folder_path.parent
    labels_folder_path = dataset_path / "labels"
    label_filename = image_path.with_suffix(".txt").name
    label_path = labels_folder_path / label_filename

    return str(label_path)

get_labels()

Parse the label file and extract label information. Each label can contain: - class: class of the object - s_xywh: scaled xywh box - xyxy: xyxy box - keypoints: list of dict(xy, state) for pose estimation tasks

Returns:

Name Type Description
labels list

List of dictionaries, each containing label information.
Source code in kano\dataset_utils.py 
 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
def get_labels(self):
    """
    Parse the label file and extract label information.
    Each label can contain:
        - class: class of the object
        - s_xywh: scaled xywh box
        - xyxy: xyxy box
        - keypoints: list of dict(xy, state) for pose estimation tasks

    Returns:
        labels (list): List of dictionaries, each containing label information.
    """
    self.label_path = self.get_label_path(self.image_path)
    labels = list()
    image_height, image_width = self.image.shape[:2]
    with open(self.label_path, "r") as file:
        for line in file:
            line = line.strip().split()
            label = {
                "class": int(line[0]),
                "s_xywh": np.array([float(x) for x in line[1:5]]),
            }
            xywh = label["s_xywh"].copy() * np.array(
                [image_width, image_height, image_width, image_height]
            )
            label["xyxy"] = xywh2xyxy(xywh)

            if self.task == "pose":
                keypoints = list()
                for start_id in range(5, len(line), 3):
                    # "state" in [0, 1, 2] with:
                    # - 0: deleted
                    # - 1: occluded
                    # - 2: visible
                    x = int(float(line[start_id]) * image_width)
                    y = int(float(line[start_id + 1]) * image_height)
                    state = int(float(line[start_id + 2]))
                    keypoints.append(
                        {
                            "xy": (x, y),
                            "state": state,
                        }
                    )
                label["keypoints"] = keypoints

            labels.append(label)
    return labels

show_annotated_image(figsize=(10, 10))

Display the annotated image with bounding boxes or skeletons drawn.

Returns:

Name Type Description
figsize tuple(int, int)

Size of the figure (width, height) in inches.
Source code in kano\dataset_utils.py 
149
150
151
152
153
154
155
156
157
def show_annotated_image(self, figsize=(10, 10)):
    """
    Display the annotated image with bounding boxes or skeletons drawn.

    Returns:
        figsize (tuple(int, int)): Size of the figure (width, height) in inches.
    """
    annotated_image = self.get_annotated_image()
    show_image(annotated_image, figsize)

show_image(figsize=(10, 10))

Display the original image.

Parameters:

Name Type Description Default
figsize tuple(int, int

Size of the figure (width, height) in inches.

 (10, 10)
Source code in kano\dataset_utils.py 
114
115
116
117
118
119
120
121
def show_image(self, figsize=(10, 10)):
    """
    Display the original image.

    Args:
        figsize (tuple(int, int)): Size of the figure (width, height) in inches.
    """
    show_image(self.image, figsize)

kano.dataset_utils.YoloDataset

Dataset class with Yolo format.

Attributes:

Name Type Description
dataset_path Path

path to the dataset folder
name str

name of the dataset (folder name)
classes list(str

names of classes in the dataset
task str

dataset usecase, must be a task in ["detect", "pose"]
Source code in kano\dataset_utils.py 
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
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
class YoloDataset:
    """
    Dataset class with Yolo format.

    Attributes:
        dataset_path (Path): path to the dataset folder
        name (str): name of the dataset (folder name)
        classes (list(str)): names of classes in the dataset
        task (str): dataset usecase, must be a task in ["detect", "pose"]
    """

    def __init__(self, dataset_path, task="detect"):
        """
        Initialize a YoloDataset object.

        Returns:
            dataset_path (str): Path to the dataset folder.
            task (str): . Dataset use case. Default is "detect".
                        Possible values: "detect", "pose".
        """
        self.dataset_path = Path(dataset_path)
        self.name = self.dataset_path.name
        self.train_folder = self.dataset_path / "train"
        self.valid_folder = self.dataset_path / "valid"
        self.test_folder = self.dataset_path / "test"
        self.classes = self.get_classes(str(self.dataset_path / "data.yaml"))
        if task not in TASKS:
            raise ValueError("Unexpected task. Please provide one of:", TASKS)
        self.task = task

    @classmethod
    def get_classes(cls, yaml_path):
        """
        Retrieve classes from a YAML file.

        Returns:
            yaml_path (str): Path to the YAML file containing class names.

        Returns:
            names (list(str)): List of class names.
        """
        with open(yaml_path, "r") as file:
            data = yaml.safe_load(file)
        return data["names"]

    def summary(self, count_box=False):
        """
        Print summary information about the dataset.
        """
        print(f"Summary dataset {self.dataset_path.name}:")
        print("- Classes: ", self.classes)

        if count_box:
            print(
                "  *Note: the box counting can take a long time depend on dataset size, please wait..."
            )
            images_paths = list()
            for folder_path in [
                self.train_folder,
                self.valid_folder,
                self.test_folder,
            ]:
                if folder_path.exists():
                    images_paths += list_files(str(folder_path / "images"))

            box_counts = {cls_name: 0 for cls_name in self.classes}
            for img_path in images_paths:
                labels = YoloImage(image_path=img_path).get_labels()
                for label in labels:
                    cls_name = self.classes[label["class"]]
                    box_counts[cls_name] += 1

            for cls_name, box_count in box_counts.items():
                print(f"  + {cls_name}: {box_count} boxes")

        print("- Subsets:")
        total_file_count = 0
        for folder_path in [
            self.train_folder,
            self.valid_folder,
            self.test_folder,
        ]:
            if folder_path.exists():
                file_count = len(list_files(str(folder_path / "images")))
                print(f"  + {folder_path.name}: {file_count} images")
                total_file_count += file_count
        print("- Total images:", total_file_count)

    @classmethod
    def _combine_classes(cls, datasets_paths):
        """
        Combine classes from multiple datasets.

        Returns:
            datasets_paths (list(str)): Paths to the dataset folders.

        Returns:
            classes (list(str)): Combined list of class names.
        """
        classes = set()

        for path in datasets_paths:
            new_classes = cls.get_classes(str(Path(path) / "data.yaml"))
            classes.update(new_classes)

        classes = list(classes)
        classes.sort()

        return classes

    @classmethod
    def _get_reindex_dict(cls, source_classes, target_classes):
        """
        Get the reindexing dictionary for mapping classes from source to target.

        Returns:
            source_classes (list(str)): Source class names.
            target_classes (list(str)): Target class names.

        Returns:
            reindex_dict (dict): Reindexing dictionary.
        """
        reindex_dict = dict()
        for i, class_name in enumerate(source_classes):
            if class_name in target_classes:
                reindex_dict[i] = target_classes.index(class_name)

        return reindex_dict

    @classmethod
    def _create_simple_yaml_file(cls, dataset_path, classes):
        """
        Create a simple YAML file containing dataset information.

        Returns:
            dataset_path (str): Path to the dataset folder.
            classes (list[str]): List of class names.
        """
        data = {
            "train": "train",
            "val": "valid",
            "test": "test",
            "names": classes,
        }

        yaml_path = Path(dataset_path) / "data.yaml"
        with open(str(yaml_path), "w") as f:
            yaml.dump(data, f)

    @classmethod
    def merge_datasets(cls, datasets_paths, merged_dataset_path):
        """
        Merge multiple datasets into one.

        Returns:
            datasets_paths (list[str]): Paths to the dataset folders to be merged.
            merged_dataset_path (str): Path to the merged dataset folder.
        """
        merged_dataset_path = Path(merged_dataset_path)
        merged_classes = cls._combine_classes(datasets_paths)
        print("Input datasets:")
        for path in datasets_paths:
            dataset = cls(path)
            dataset.summary()
            classes = dataset.classes
            reindex_dict = cls._get_reindex_dict(classes, merged_classes)

            for subset_path in [
                dataset.train_folder,
                dataset.valid_folder,
                dataset.test_folder,
            ]:
                if subset_path.exists():
                    target_subset_path = merged_dataset_path / subset_path.name
                    create_folder(target_subset_path / "images")
                    create_folder(target_subset_path / "labels")
                    images_paths = list_files(str(subset_path / "images"))
                    for image_path in images_paths:
                        yolo_image = YoloImage(image_path)
                        yolo_image.copy_to(
                            target_subset_path,
                            dataset.name + "_",
                            reindex_dict,
                        )

        cls._create_simple_yaml_file(str(merged_dataset_path), merged_classes)

        print("Merged dataset:")
        dataset = cls(str(merged_dataset_path))
        dataset.summary()

    def split(self, splitted_dataset_path, ratios=[0.9]):
        """
        Split the dataset into train, validation, and test subsets.

        Returns:
            splitted_dataset_path (str): Path to the folder where the splitted dataset will be saved.
            ratios (list[float]): Ratios for train, validation, and test subsets. Default is [0.9].
        """
        self.summary()

        images_paths = list()
        for folder_path in [
            self.train_folder,
            self.valid_folder,
            self.test_folder,
        ]:
            if folder_path.exists():
                images_paths += list_files(str(folder_path / "images"))

        random.shuffle(images_paths)

        total_paths = len(images_paths)

        train_count = int(total_paths * ratios[0])
        train_paths = images_paths[:train_count]

        if len(ratios) == 1:
            valid_paths = images_paths[train_count:]
            test_paths = list()
        else:
            valid_count = int(total_paths * ratios[1])
            valid_paths = images_paths[train_count : train_count + valid_count]
            test_paths = images_paths[train_count + valid_count :]

        splitted_dataset_path = Path(splitted_dataset_path)

        subsets = [
            ("train", train_paths),
            ("valid", valid_paths),
            ("test", test_paths),
        ]
        for subset_name, paths in subsets:
            for path in paths:
                yolo_image = YoloImage(path)
                old_subset_name = Path(path).parent.parent.name
                target_folder_path = splitted_dataset_path / subset_name
                create_folder(target_folder_path / "images")
                create_folder(target_folder_path / "labels")
                yolo_image.copy_to(
                    target_folder_path, f"{self.name}_{old_subset_name}_"
                )

        self._create_simple_yaml_file(str(splitted_dataset_path), self.classes)
        YoloDataset(str(splitted_dataset_path)).summary()

    def rename_classes(self, renamed_dataset_path, renaming_dict):
        """
        Rename classes in the dataset.

        Returns:
            renamed_dataset_path (str): Path to the folder where the renamed dataset will be saved.
            renaming_dict (dict): Dictionary mapping original class names to new class names.
                To remove classes, set the classes' values to None.
        """

        self.summary()

        target_classes = [
            class_name
            for class_name in renaming_dict.values()
            if class_name is not None
        ]

        for class_name in self.classes:
            if class_name not in renaming_dict.keys():
                target_classes.append(class_name)

        target_classes = list(set(target_classes))
        target_classes.sort()

        print("Classes after renaming:", target_classes)

        reindex_dict = dict()

        for i, class_name in enumerate(self.classes):
            renamed_class = class_name
            if class_name in renaming_dict:
                renamed_class = renaming_dict[class_name]

            if renamed_class is None:
                reindex_dict[i] = None
            else:
                reindex_dict[i] = target_classes.index(renamed_class)

        renamed_dataset_path = Path(renamed_dataset_path)
        for subset_name, folder_path in [
            ("train", self.train_folder),
            ("valid", self.valid_folder),
            ("test", self.test_folder),
        ]:
            if folder_path.exists():
                images_paths = list_files(folder_path / "images")
                for path in images_paths:
                    yolo_image = YoloImage(path)
                    target_folder_path = renamed_dataset_path / subset_name
                    create_folder(target_folder_path / "images")
                    create_folder(target_folder_path / "labels")
                    yolo_image.copy_to(
                        target_folder_path, f"{self.name}_", reindex_dict
                    )

        self._create_simple_yaml_file(
            str(renamed_dataset_path), target_classes
        )
        YoloDataset(str(renamed_dataset_path)).summary()

    def show_sample(self, figsize=(10, 10)):
        """
        Show a sample of annotated images from the dataset.

        Returns:
            figsize (tuple(int, int)): Size of the figure (width, height) in inches.
        """

        images_paths = list()
        for folder_path in [
            self.train_folder,
            self.valid_folder,
            self.test_folder,
        ]:
            if folder_path.exists():
                images_paths += list_files(str(folder_path / "images"))

        random.shuffle(images_paths)

        labels_dict = {
            i: class_name for i, class_name in enumerate(self.classes)
        }
        annotated_images = list()
        if len(images_paths) > 9:
            for i in range(3):
                annotated_images.append(list())
                for j in range(3):
                    yolo_image = YoloImage(
                        images_paths[i * 3 + j], labels_dict, self.task
                    )
                    annotated_images[i].append(
                        yolo_image.get_annotated_image()
                    )
        else:
            total_images = min(3, len(images_paths))
            for i in range(total_images):
                yolo_image = YoloImage(images_paths[i], labels_dict, self.task)
                annotated_images.append(yolo_image.get_annotated_image())

        concatenated_images = concatenate_images(annotated_images)

        show_image(concatenated_images, figsize=figsize)

    def number_filenames(self, numbered_dataset_path):
        """
        Number filenames of images in the dataset.

        Returns:
            numbered_dataset_path (str): Path to the folder where the numbered dataset will be saved.
        """
        self.summary()

        total_images = 0

        for folder_path in [
            self.train_folder,
            self.valid_folder,
            self.test_folder,
        ]:
            if folder_path.exists():
                total_images += len(list_files(str(folder_path / "images")))

        n_digits = len(str(total_images))
        i = 0

        subsets = [
            ("train", self.train_folder),
            ("valid", self.valid_folder),
            ("test", self.test_folder),
        ]
        numbered_dataset_path = Path(numbered_dataset_path)
        for subset_name, folder_path in subsets:
            if folder_path.exists():
                images_paths = list_files(str(folder_path / "images"))
                for path in images_paths:
                    yolo_image = YoloImage(path)
                    target_folder_path = numbered_dataset_path / subset_name
                    create_folder(target_folder_path / "images")
                    create_folder(target_folder_path / "labels")
                    yolo_image.copy_to(
                        target_folder_path, target_stem=str(i).zfill(n_digits)
                    )
                    i += 1

        self._create_simple_yaml_file(str(numbered_dataset_path), self.classes)
        YoloDataset(str(numbered_dataset_path)).summary()

__init__(dataset_path, task='detect')

Initialize a YoloDataset object.

Returns:

Name Type Description
dataset_path str

Path to the dataset folder.
task str

. Dataset use case. Default is "detect". Possible values: "detect", "pose".
Source code in kano\dataset_utils.py 
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
def __init__(self, dataset_path, task="detect"):
    """
    Initialize a YoloDataset object.

    Returns:
        dataset_path (str): Path to the dataset folder.
        task (str): . Dataset use case. Default is "detect".
                    Possible values: "detect", "pose".
    """
    self.dataset_path = Path(dataset_path)
    self.name = self.dataset_path.name
    self.train_folder = self.dataset_path / "train"
    self.valid_folder = self.dataset_path / "valid"
    self.test_folder = self.dataset_path / "test"
    self.classes = self.get_classes(str(self.dataset_path / "data.yaml"))
    if task not in TASKS:
        raise ValueError("Unexpected task. Please provide one of:", TASKS)
    self.task = task

get_classes(yaml_path) classmethod

Retrieve classes from a YAML file.

Returns:

Name Type Description
yaml_path str

Path to the YAML file containing class names.

Returns:

Name Type Description
names list(str)

List of class names.
Source code in kano\dataset_utils.py 
250
251
252
253
254
255
256
257
258
259
260
261
262
263
@classmethod
def get_classes(cls, yaml_path):
    """
    Retrieve classes from a YAML file.

    Returns:
        yaml_path (str): Path to the YAML file containing class names.

    Returns:
        names (list(str)): List of class names.
    """
    with open(yaml_path, "r") as file:
        data = yaml.safe_load(file)
    return data["names"]

merge_datasets(datasets_paths, merged_dataset_path) classmethod

Merge multiple datasets into one.

Returns:

Name Type Description
datasets_paths list[str]

Paths to the dataset folders to be merged.
merged_dataset_path str

Path to the merged dataset folder.
Source code in kano\dataset_utils.py 
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
@classmethod
def merge_datasets(cls, datasets_paths, merged_dataset_path):
    """
    Merge multiple datasets into one.

    Returns:
        datasets_paths (list[str]): Paths to the dataset folders to be merged.
        merged_dataset_path (str): Path to the merged dataset folder.
    """
    merged_dataset_path = Path(merged_dataset_path)
    merged_classes = cls._combine_classes(datasets_paths)
    print("Input datasets:")
    for path in datasets_paths:
        dataset = cls(path)
        dataset.summary()
        classes = dataset.classes
        reindex_dict = cls._get_reindex_dict(classes, merged_classes)

        for subset_path in [
            dataset.train_folder,
            dataset.valid_folder,
            dataset.test_folder,
        ]:
            if subset_path.exists():
                target_subset_path = merged_dataset_path / subset_path.name
                create_folder(target_subset_path / "images")
                create_folder(target_subset_path / "labels")
                images_paths = list_files(str(subset_path / "images"))
                for image_path in images_paths:
                    yolo_image = YoloImage(image_path)
                    yolo_image.copy_to(
                        target_subset_path,
                        dataset.name + "_",
                        reindex_dict,
                    )

    cls._create_simple_yaml_file(str(merged_dataset_path), merged_classes)

    print("Merged dataset:")
    dataset = cls(str(merged_dataset_path))
    dataset.summary()

number_filenames(numbered_dataset_path)

Number filenames of images in the dataset.

Returns:

Name Type Description
numbered_dataset_path str

Path to the folder where the numbered dataset will be saved.
Source code in kano\dataset_utils.py 
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
def number_filenames(self, numbered_dataset_path):
    """
    Number filenames of images in the dataset.

    Returns:
        numbered_dataset_path (str): Path to the folder where the numbered dataset will be saved.
    """
    self.summary()

    total_images = 0

    for folder_path in [
        self.train_folder,
        self.valid_folder,
        self.test_folder,
    ]:
        if folder_path.exists():
            total_images += len(list_files(str(folder_path / "images")))

    n_digits = len(str(total_images))
    i = 0

    subsets = [
        ("train", self.train_folder),
        ("valid", self.valid_folder),
        ("test", self.test_folder),
    ]
    numbered_dataset_path = Path(numbered_dataset_path)
    for subset_name, folder_path in subsets:
        if folder_path.exists():
            images_paths = list_files(str(folder_path / "images"))
            for path in images_paths:
                yolo_image = YoloImage(path)
                target_folder_path = numbered_dataset_path / subset_name
                create_folder(target_folder_path / "images")
                create_folder(target_folder_path / "labels")
                yolo_image.copy_to(
                    target_folder_path, target_stem=str(i).zfill(n_digits)
                )
                i += 1

    self._create_simple_yaml_file(str(numbered_dataset_path), self.classes)
    YoloDataset(str(numbered_dataset_path)).summary()

rename_classes(renamed_dataset_path, renaming_dict)

Rename classes in the dataset.

Returns:

Name Type Description
renamed_dataset_path str

Path to the folder where the renamed dataset will be saved.
renaming_dict dict

Dictionary mapping original class names to new class names. To remove classes, set the classes' values to None.
Source code in kano\dataset_utils.py 
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
def rename_classes(self, renamed_dataset_path, renaming_dict):
    """
    Rename classes in the dataset.

    Returns:
        renamed_dataset_path (str): Path to the folder where the renamed dataset will be saved.
        renaming_dict (dict): Dictionary mapping original class names to new class names.
            To remove classes, set the classes' values to None.
    """

    self.summary()

    target_classes = [
        class_name
        for class_name in renaming_dict.values()
        if class_name is not None
    ]

    for class_name in self.classes:
        if class_name not in renaming_dict.keys():
            target_classes.append(class_name)

    target_classes = list(set(target_classes))
    target_classes.sort()

    print("Classes after renaming:", target_classes)

    reindex_dict = dict()

    for i, class_name in enumerate(self.classes):
        renamed_class = class_name
        if class_name in renaming_dict:
            renamed_class = renaming_dict[class_name]

        if renamed_class is None:
            reindex_dict[i] = None
        else:
            reindex_dict[i] = target_classes.index(renamed_class)

    renamed_dataset_path = Path(renamed_dataset_path)
    for subset_name, folder_path in [
        ("train", self.train_folder),
        ("valid", self.valid_folder),
        ("test", self.test_folder),
    ]:
        if folder_path.exists():
            images_paths = list_files(folder_path / "images")
            for path in images_paths:
                yolo_image = YoloImage(path)
                target_folder_path = renamed_dataset_path / subset_name
                create_folder(target_folder_path / "images")
                create_folder(target_folder_path / "labels")
                yolo_image.copy_to(
                    target_folder_path, f"{self.name}_", reindex_dict
                )

    self._create_simple_yaml_file(
        str(renamed_dataset_path), target_classes
    )
    YoloDataset(str(renamed_dataset_path)).summary()

show_sample(figsize=(10, 10))

Show a sample of annotated images from the dataset.

Returns:

Name Type Description
figsize tuple(int, int)

Size of the figure (width, height) in inches.
Source code in kano\dataset_utils.py 
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
def show_sample(self, figsize=(10, 10)):
    """
    Show a sample of annotated images from the dataset.

    Returns:
        figsize (tuple(int, int)): Size of the figure (width, height) in inches.
    """

    images_paths = list()
    for folder_path in [
        self.train_folder,
        self.valid_folder,
        self.test_folder,
    ]:
        if folder_path.exists():
            images_paths += list_files(str(folder_path / "images"))

    random.shuffle(images_paths)

    labels_dict = {
        i: class_name for i, class_name in enumerate(self.classes)
    }
    annotated_images = list()
    if len(images_paths) > 9:
        for i in range(3):
            annotated_images.append(list())
            for j in range(3):
                yolo_image = YoloImage(
                    images_paths[i * 3 + j], labels_dict, self.task
                )
                annotated_images[i].append(
                    yolo_image.get_annotated_image()
                )
    else:
        total_images = min(3, len(images_paths))
        for i in range(total_images):
            yolo_image = YoloImage(images_paths[i], labels_dict, self.task)
            annotated_images.append(yolo_image.get_annotated_image())

    concatenated_images = concatenate_images(annotated_images)

    show_image(concatenated_images, figsize=figsize)

split(splitted_dataset_path, ratios=[0.9])

Split the dataset into train, validation, and test subsets.

Returns:

Name Type Description
splitted_dataset_path str

Path to the folder where the splitted dataset will be saved.
ratios list[float]

Ratios for train, validation, and test subsets. Default is [0.9].
Source code in kano\dataset_utils.py 
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
def split(self, splitted_dataset_path, ratios=[0.9]):
    """
    Split the dataset into train, validation, and test subsets.

    Returns:
        splitted_dataset_path (str): Path to the folder where the splitted dataset will be saved.
        ratios (list[float]): Ratios for train, validation, and test subsets. Default is [0.9].
    """
    self.summary()

    images_paths = list()
    for folder_path in [
        self.train_folder,
        self.valid_folder,
        self.test_folder,
    ]:
        if folder_path.exists():
            images_paths += list_files(str(folder_path / "images"))

    random.shuffle(images_paths)

    total_paths = len(images_paths)

    train_count = int(total_paths * ratios[0])
    train_paths = images_paths[:train_count]

    if len(ratios) == 1:
        valid_paths = images_paths[train_count:]
        test_paths = list()
    else:
        valid_count = int(total_paths * ratios[1])
        valid_paths = images_paths[train_count : train_count + valid_count]
        test_paths = images_paths[train_count + valid_count :]

    splitted_dataset_path = Path(splitted_dataset_path)

    subsets = [
        ("train", train_paths),
        ("valid", valid_paths),
        ("test", test_paths),
    ]
    for subset_name, paths in subsets:
        for path in paths:
            yolo_image = YoloImage(path)
            old_subset_name = Path(path).parent.parent.name
            target_folder_path = splitted_dataset_path / subset_name
            create_folder(target_folder_path / "images")
            create_folder(target_folder_path / "labels")
            yolo_image.copy_to(
                target_folder_path, f"{self.name}_{old_subset_name}_"
            )

    self._create_simple_yaml_file(str(splitted_dataset_path), self.classes)
    YoloDataset(str(splitted_dataset_path)).summary()

summary(count_box=False)

Print summary information about the dataset.

Source code in kano\dataset_utils.py 
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
def summary(self, count_box=False):
    """
    Print summary information about the dataset.
    """
    print(f"Summary dataset {self.dataset_path.name}:")
    print("- Classes: ", self.classes)

    if count_box:
        print(
            "  *Note: the box counting can take a long time depend on dataset size, please wait..."
        )
        images_paths = list()
        for folder_path in [
            self.train_folder,
            self.valid_folder,
            self.test_folder,
        ]:
            if folder_path.exists():
                images_paths += list_files(str(folder_path / "images"))

        box_counts = {cls_name: 0 for cls_name in self.classes}
        for img_path in images_paths:
            labels = YoloImage(image_path=img_path).get_labels()
            for label in labels:
                cls_name = self.classes[label["class"]]
                box_counts[cls_name] += 1

        for cls_name, box_count in box_counts.items():
            print(f"  + {cls_name}: {box_count} boxes")

    print("- Subsets:")
    total_file_count = 0
    for folder_path in [
        self.train_folder,
        self.valid_folder,
        self.test_folder,
    ]:
        if folder_path.exists():
            file_count = len(list_files(str(folder_path / "images")))
            print(f"  + {folder_path.name}: {file_count} images")
            total_file_count += file_count
    print("- Total images:", total_file_count)