Skip to content

Versions

Version

Class representing a Roboflow dataset version.

Source code in roboflow/core/version.py
 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
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
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
class Version:
    """
    Class representing a Roboflow dataset version.
    """

    model: Optional[InferenceModel]

    def __init__(
        self,
        version_dict,
        type,
        api_key,
        name,
        version,
        model_format,
        local: Optional[str],
        workspace,
        project,
        public,
        colors=None,
    ):
        """
        Initialize a Version object.
        """
        if api_key:
            self.__api_key = api_key
            self.name = name
            self.version = unwrap_version_id(version_id=version)
            self.type = type
            self.augmentation = version_dict["augmentation"]
            self.created = version_dict["created"]
            self.id = version_dict["id"]
            self.images = version_dict["images"]
            self.preprocessing = version_dict["preprocessing"]
            self.splits = version_dict["splits"]
            self.model_format = model_format
            self.workspace = workspace
            self.project = project
            self.public = public
            self.colors = {} if colors is None else colors

            self.colors = colors
            if "exports" in version_dict.keys():
                self.exports = version_dict["exports"]
            else:
                self.exports = []

            version_without_workspace = os.path.basename(str(version))

            response = requests.get(f"{API_URL}/{workspace}/{project}/{self.version}?api_key={self.__api_key}")
            if response.ok:
                version_info = response.json()["version"]
                has_model = bool(version_info.get("train", {}).get("model"))
            else:
                has_model = False

            if not has_model:
                self.model = None
            elif self.type == TYPE_OBJECT_DETECTION:
                self.model = ObjectDetectionModel(
                    self.__api_key,
                    self.id,
                    self.name,
                    version_without_workspace,
                    local=local,
                    colors=self.colors,
                    preprocessing=self.preprocessing,
                )
            elif self.type == TYPE_CLASSICATION:
                self.model = ClassificationModel(
                    self.__api_key,
                    self.id,
                    self.name,
                    version_without_workspace,
                    local=local,
                    colors=self.colors,
                    preprocessing=self.preprocessing,
                )
            elif self.type == TYPE_INSTANCE_SEGMENTATION:
                self.model = InstanceSegmentationModel(
                    self.__api_key,
                    self.id,
                    colors=self.colors,
                    preprocessing=self.preprocessing,
                    local=local,
                )
            elif self.type == TYPE_SEMANTIC_SEGMENTATION:
                self.model = SemanticSegmentationModel(self.__api_key, self.id)
            elif self.type == TYPE_KEYPOINT_DETECTION:
                self.model = KeypointDetectionModel(self.__api_key, self.id, version=version_without_workspace)
            else:
                self.model = None

        elif DEMO_KEYS:
            api_key = DEMO_KEYS[0]
            if api_key == "coco-128-sample":
                self.__api_key = api_key
                self.model_format = model_format
                self.name = "coco-128"
                self.version = "1"
            else:
                self.__api_key = api_key
                self.model_format = model_format
                self.name = "chess-pieces-new"
                self.version = "23"
                self.id = "joseph-nelson/chess-pieces-new"

    def __check_if_generating(self):
        # check Roboflow API to see if this version is still generating

        url = f"{API_URL}/{self.workspace}/{self.project}/{self.version}?nocache=true"
        response = requests.get(url, params={"api_key": self.__api_key})
        response.raise_for_status()
        if response.json()["version"]["progress"] is None:
            progress = 0.0
        else:
            progress = float(response.json()["version"]["progress"])

        return response.json()["version"]["generating"], progress

    def __wait_if_generating(self, recurse=False):
        # checks if a given version is still in the progress of generating

        still_generating, progress = self.__check_if_generating()

        if still_generating:
            progress_message = "Generating version still in progress. Progress: " + str(round(progress * 100, 2)) + "%"
            sys.stdout.write("\r" + progress_message)
            sys.stdout.flush()
            time.sleep(5)
            return self.__wait_if_generating(recurse=True)

        else:
            if recurse:
                sys.stdout.write("\n")
                sys.stdout.flush()
            return

    def download(self, model_format=None, location=None, overwrite: bool = False):
        """
        Download and extract a ZIP of a version's dataset in a given format

        :param model_format: A format to use for downloading
        :param location: An optional path for saving the file
        :param overwrite: An optional flag to prevent dataset overwrite when dataset is already downloaded

        Args:
            model_format (str): A format to use for downloading
            location (str): An optional path for saving the file
            overwrite (bool): An optional flag to overwrite an existing dataset if the dataset has already downloaded

        Returns:
            Dataset Object

        Raises:
            RuntimeError: If the Roboflow API returns an error with a helpful JSON body
            HTTPError: If the Network/Roboflow API fails and does not return JSON
        """  # noqa: E501 // docs

        self.__wait_if_generating()

        model_format = self.__get_format_identifier(model_format)

        if model_format not in self.exports:
            self.export(model_format)

        # if model_format is not in

        if location is None:
            location = self.__get_download_location()
        if os.path.exists(location) and not overwrite:
            return Dataset(self.name, self.version, model_format, os.path.abspath(location))

        if self.__api_key == "coco-128-sample":
            link = "https://app.roboflow.com/ds/n9QwXwUK42?key=NnVCe2yMxP"
        else:
            url = self.__get_download_url(model_format)
            response = requests.get(url, params={"api_key": self.__api_key})
            if response.status_code == 200:
                link = response.json()["export"]["link"]
            else:
                try:
                    raise RuntimeError(response.json())
                except json.JSONDecodeError:
                    response.raise_for_status()

        self.__download_zip(link, location, model_format)
        self.__extract_zip(location, model_format)
        self.__reformat_yaml(location, model_format)  # TODO: is roboflow-python a place to be munging yaml files?

        return Dataset(self.name, self.version, model_format, os.path.abspath(location))

    def export(self, model_format=None):
        """
        Ask the Roboflow API to generate a version's dataset in a given format so that it can be downloaded via the `download()` method.

        The export will be asynchronously generated and available for download after some amount of seconds - depending on dataset size.

        Args:
            model_format (str): A format to use for downloading

        Returns:
            True

        Raises:
            RuntimeError: If the Roboflow API returns an error with a helpful JSON body
            HTTPError: If the Network/Roboflow API fails and does not return JSON
        """  # noqa: E501 // docs

        model_format = self.__get_format_identifier(model_format)

        self.__wait_if_generating()

        url = self.__get_download_url(model_format)
        response = requests.get(url, params={"api_key": self.__api_key})
        if not response.ok:
            try:
                raise RuntimeError(response.json())
            except json.JSONDecodeError:
                response.raise_for_status()

        # the rest api returns 202 if the export is still in progress
        if response.status_code == 202:
            status_code_check = 202
            while status_code_check == 202:
                time.sleep(1)
                response = requests.get(url, params={"api_key": self.__api_key})
                status_code_check = response.status_code
                if status_code_check == 202:
                    progress = response.json()["progress"]
                    progress_message = (
                        "Exporting format " + model_format + " in progress : " + str(round(progress * 100, 2)) + "%"
                    )
                    sys.stdout.write("\r" + progress_message)
                    sys.stdout.flush()

        if response.status_code == 200:
            sys.stdout.write("\n")
            print("\r" + "Version export complete for " + model_format + " format")
            sys.stdout.flush()
            return True
        else:
            try:
                raise RuntimeError(response.json())
            except json.JSONDecodeError:
                response.raise_for_status()

    def train(self, speed=None, checkpoint=None, plot_in_notebook=False) -> InferenceModel:
        """
        Ask the Roboflow API to train a previously exported version's dataset.

        Args:
            speed: Whether to train quickly or accurately. Note: accurate training is a paid feature. Default speed is `fast`.
            checkpoint: A string representing the checkpoint to use while training
            plot: Whether to plot the training results. Default is `False`.

        Returns:
            An instance of the trained model class

        Raises:
            RuntimeError: If the Roboflow API returns an error with a helpful JSON body
            HTTPError: If the Network/Roboflow API fails and does not return JSON
        """  # noqa: E501 // docs

        self.__wait_if_generating()

        train_model_format = "yolov5pytorch"

        if self.type == TYPE_CLASSICATION:
            train_model_format = "folder"

        if self.type == TYPE_INSTANCE_SEGMENTATION:
            train_model_format = "yolov5pytorch"

        if self.type == TYPE_SEMANTIC_SEGMENTATION:
            train_model_format = "png-mask-semantic"

        # if classification
        if train_model_format not in self.exports:
            self.export(train_model_format)

        workspace, project, *_ = self.id.rsplit("/")
        url = f"{API_URL}/{workspace}/{project}/{self.version}/train"

        data = {}
        if speed:
            data["speed"] = speed

        if checkpoint:
            data["checkpoint"] = checkpoint

        write_line("Reaching out to Roboflow to start training...")

        response = requests.post(url, json=data, params={"api_key": self.__api_key})
        if not response.ok:
            try:
                raise RuntimeError(response.json())
            except json.JSONDecodeError:
                response.raise_for_status()

        status = "training"

        if plot_in_notebook:
            from IPython.display import clear_output
            from matplotlib import pyplot as plt

            def live_plot(epochs, mAP, loss, title=""):
                clear_output(wait=True)

                plt.subplot(2, 1, 1)
                plt.plot(epochs, mAP, "#00FFCE")
                plt.title(title)
                plt.ylabel("mAP")

                plt.subplot(2, 1, 2)
                plt.plot(epochs, loss, "#A351FB")
                plt.xlabel("epochs")
                plt.ylabel("loss")
                plt.show()

        first_graph_write = False
        previous_epochs: Union[np.ndarray, list] = []
        num_machine_spin_dots = []

        while status == "training" or status == "running":
            url = f"{API_URL}/{self.workspace}/{self.project}/{self.version}?nocache=true"
            response = requests.get(url, params={"api_key": self.__api_key})
            response.raise_for_status()
            version = response.json()["version"]
            if "models" in version.keys():
                models = version["models"]
            else:
                models = {}

            if "train" in version.keys():
                if "results" in version["train"].keys():
                    status = "finished"
                    break
                if "status" in version["train"].keys():
                    if version["train"]["status"] == "failed":
                        write_line(line="Training failed")
                        break

            epochs: Union[np.ndarray, list]
            mAP: Union[np.ndarray, list]
            loss: Union[np.ndarray, list]

            if "roboflow-train" in models.keys():
                import numpy as np

                # training has started
                epochs = np.array([int(epoch["epoch"]) for epoch in models["roboflow-train"]["epochs"]])
                mAP = np.array([float(epoch["mAP"]) for epoch in models["roboflow-train"]["epochs"]])
                loss = np.array(
                    [
                        sum(float(epoch[key]) for key in ["box_loss", "class_loss", "obj_loss"] if key in epoch)
                        for epoch in models["roboflow-train"]["epochs"]
                    ]
                )

                title = "Training in Progress"
                # plottling logic
            else:
                num_machine_spin_dots.append(".")
                if len(num_machine_spin_dots) > 5:
                    num_machine_spin_dots = ["."]
                title = "Training Machine Spinning Up" + "".join(num_machine_spin_dots)

                epochs = []
                mAP = []
                loss = []

            if (len(epochs) > len(previous_epochs)) or (len(epochs) == 0):
                if plot_in_notebook:
                    live_plot(epochs, mAP, loss, title)
                else:
                    if len(epochs) > 0:
                        title = (
                            title + ": Epoch: " + str(epochs[-1]) + " mAP: " + str(mAP[-1]) + " loss: " + str(loss[-1])
                        )
                    if not first_graph_write:
                        write_line(title)
                        first_graph_write = True

            previous_epochs = copy.deepcopy(epochs)

            time.sleep(5)

        if not self.model:
            if self.type == TYPE_OBJECT_DETECTION:
                self.model = ObjectDetectionModel(
                    self.__api_key,
                    self.id,
                    self.name,
                    self.version,
                    colors=self.colors,
                    preprocessing=self.preprocessing,
                )
            elif self.type == TYPE_CLASSICATION:
                self.model = ClassificationModel(
                    self.__api_key,
                    self.id,
                    self.name,
                    self.version,
                    colors=self.colors,
                    preprocessing=self.preprocessing,
                )
            elif self.type == TYPE_INSTANCE_SEGMENTATION:
                self.model = InstanceSegmentationModel(
                    self.__api_key,
                    self.id,
                    colors=self.colors,
                    preprocessing=self.preprocessing,
                )
            elif self.type == TYPE_SEMANTIC_SEGMENTATION:
                self.model = SemanticSegmentationModel(self.__api_key, self.id)
            elif self.type == TYPE_KEYPOINT_DETECTION:
                self.model = KeypointDetectionModel(self.__api_key, self.id, version=self.version)
            else:
                raise ValueError(f"Unsupported model type: {self.type}")

        # return the model object
        assert self.model
        return self.model

    # @warn_for_wrong_dependencies_versions([("ultralytics", "==", "8.0.196")])
    def deploy(self, model_type: str, model_path: str, filename: str = "weights/best.pt") -> None:
        """Uploads provided weights file to Roboflow.

        Args:
            model_type (str): The type of the model to be deployed.
            model_path (str): File path to the model weights to be uploaded.
            filename (str, optional): The name of the weights file. Defaults to "weights/best.pt".
        """
        model_type = normalize_yolo_model_type(model_type)
        zip_file_name = process(model_type, model_path, filename)

        if zip_file_name is None:
            raise RuntimeError("Failed to process model")

        self._upload_zip(model_type, model_path, zip_file_name)

    def _upload_zip(self, model_type: str, model_path: str, model_file_name: str):
        res = requests.get(
            f"{API_URL}/{self.workspace}/{self.project}/{self.version}"
            f"/uploadModel?api_key={self.__api_key}&modelType={model_type}&nocache=true"
        )
        try:
            if res.status_code == 429:
                raise RuntimeError(
                    "This version already has a trained model. Please generate and"
                    " train a new version in order to upload model to Roboflow."
                )
            else:
                res.raise_for_status()
        except Exception as e:
            print(f"An error occured when getting the model upload URL: {e}")
            return

        res = requests.put(
            res.json()["url"],
            data=open(os.path.join(model_path, model_file_name), "rb"),
        )
        try:
            res.raise_for_status()

            if self.public:
                print(
                    f"View the status of your deployment at: {APP_URL}/{self.workspace}/{self.project}/{self.version}"
                )
                print(
                    "Share your model with the world at:"
                    f" {UNIVERSE_URL}/{self.workspace}/{self.project}/"
                    f"model/{self.version}"
                )
            else:
                print(
                    f"View the status of your deployment at: {APP_URL}/{self.workspace}/{self.project}/{self.version}"
                )

        except Exception as e:
            print(f"An error occured when uploading the model: {e}")

    def __download_zip(self, link, location, format):
        """
        Download a dataset's zip file from the given URL and save it in the desired location

        Args:
            link (str): link the URL of the remote zip file
            location (str): filepath of the data directory to save the zip file to
            format (str): the format identifier string
        """  # noqa: E501 // docs
        if not os.path.exists(location):
            os.makedirs(location)

        def bar_progress(current, total, width=80):
            progress_message = (
                f"Downloading Dataset Version Zip in {location} to {format}: "
                f"{current / total * 100:.0f}% [{current} / {total}] bytes"
            )
            sys.stdout.write("\r" + progress_message)
            sys.stdout.flush()

        try:
            response = requests.get(link, stream=True)

            # write the zip file to the desired location
            with open(location + "/roboflow.zip", "wb") as f:
                total_length = int(response.headers.get("content-length"))  # type: ignore[arg-type]
                desc = None if TQDM_DISABLE else f"Downloading Dataset Version Zip in {location} to {format}:"
                for chunk in tqdm(
                    response.iter_content(chunk_size=1024),
                    desc=desc,
                    total=int(total_length / 1024) + 1,
                ):
                    if chunk:
                        f.write(chunk)
                        f.flush()

        except Exception as e:
            print(f"Error when trying to download dataset @ {link}")
            raise e
        sys.stdout.write("\n")
        sys.stdout.flush()

    def __extract_zip(self, location, format):
        """
        Extracts the contents of a downloaded ZIP file and then deletes the zipped file.

        Args:
            location (str): filepath of the data directory that contains the ZIP file
            format (str): the format identifier string

        Raises:
            RuntimeError: If there is an error unzipping the file
        """  # noqa: E501 // docs
        desc = None if TQDM_DISABLE else f"Extracting Dataset Version Zip to {location} in {format}:"
        with zipfile.ZipFile(location + "/roboflow.zip", "r") as zip_ref:
            for member in tqdm(
                zip_ref.infolist(),
                desc=desc,
            ):
                try:
                    zip_ref.extract(member, location)
                except zipfile.error:
                    raise RuntimeError("Error unzipping download")

        os.remove(location + "/roboflow.zip")

    def __get_download_location(self):
        """
        Get the local path to save a downloaded dataset to

        Returns:
            str: the local path
        """
        version_slug = self.name.replace(" ", "-")
        filename = f"{version_slug}-{self.version}"

        directory = os.environ.get("DATASET_DIRECTORY")
        if directory:
            return f"{directory}/{filename}"

        return filename

    def __get_download_url(self, format):
        """
        Get the Roboflow API URL for downloading (and exporting downloadable zips)

        Args:
            format (str): the format identifier string

        Returns:
            str: the Roboflow API URL
        """
        workspace, project, *_ = self.id.rsplit("/")
        return f"{API_URL}/{workspace}/{project}/{self.version}/{format}"

    def __get_format_identifier(self, format):
        """
        If `format` is none, fall back to the instance's `model_format` value.

        If a human readable format name was passed, return the identifier that should be used for Roboflow API calls

        Otherwise, assume that the passed in format is also the identifier

        Args:
            format (str): a human readable format string

        Returns:
            str: format identifier string
        """  # noqa: E501 // docs
        if not format:
            format = self.model_format

        if not format:
            raise RuntimeError(
                "You must pass a format argument to version.download() or define a model in your Roboflow object"
            )

        friendly_formats = {"yolov5": "yolov5pytorch", "yolov7": "yolov7pytorch"}

        return friendly_formats.get(format, format)

    def __reformat_yaml(self, location: str, format: str):
        """
        Certain formats seem to require reformatting the downloaded YAML.

        Args:
            location (str): filepath of the data directory that contains the yaml file
            format (str): the format identifier string
        """  # noqa: E501 // docs
        data_path = os.path.join(location, "data.yaml")

        def data_yaml_callback(content: dict) -> dict:
            if format == "mt-yolov6":
                content["train"] = location + content["train"].lstrip(".")
                content["val"] = location + content["val"].lstrip(".")
                content["test"] = location + content["test"].lstrip(".")
            if format in ["yolov5pytorch", "yolov7pytorch"]:
                content["train"] = location + content["train"].lstrip("..")
                content["val"] = location + content["val"].lstrip("..")
            try:
                # get_wrong_dependencies_versions raises exception if ultralytics is not installed at all  # noqa: E501 // docs
                if format == "yolov8" and not get_wrong_dependencies_versions(
                    dependencies_versions=[("ultralytics", "==", "8.0.196")]
                ):
                    content["train"] = "train/images"
                    content["val"] = "valid/images"
                    content["test"] = "test/images"
            except ModuleNotFoundError:
                pass
            return content

        if format in ["yolov5pytorch", "mt-yolov6", "yolov7pytorch", "yolov8", "yolov9"]:
            amend_data_yaml(path=data_path, callback=data_yaml_callback)

    def __str__(self):
        """
        String representation of version object.
        """
        json_value = {
            "name": self.name,
            "type": self.type,
            "version": self.version,
            "augmentation": self.augmentation,
            "created": self.created,
            "preprocessing": self.preprocessing,
            "splits": self.splits,
            "workspace": self.workspace,
        }
        return json.dumps(json_value, indent=2)

__download_zip(link, location, format)

Download a dataset's zip file from the given URL and save it in the desired location

Parameters:

Name Type Description Default
link str

link the URL of the remote zip file

required
location str

filepath of the data directory to save the zip file to

required
format str

the format identifier string

required
Source code in roboflow/core/version.py
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
def __download_zip(self, link, location, format):
    """
    Download a dataset's zip file from the given URL and save it in the desired location

    Args:
        link (str): link the URL of the remote zip file
        location (str): filepath of the data directory to save the zip file to
        format (str): the format identifier string
    """  # noqa: E501 // docs
    if not os.path.exists(location):
        os.makedirs(location)

    def bar_progress(current, total, width=80):
        progress_message = (
            f"Downloading Dataset Version Zip in {location} to {format}: "
            f"{current / total * 100:.0f}% [{current} / {total}] bytes"
        )
        sys.stdout.write("\r" + progress_message)
        sys.stdout.flush()

    try:
        response = requests.get(link, stream=True)

        # write the zip file to the desired location
        with open(location + "/roboflow.zip", "wb") as f:
            total_length = int(response.headers.get("content-length"))  # type: ignore[arg-type]
            desc = None if TQDM_DISABLE else f"Downloading Dataset Version Zip in {location} to {format}:"
            for chunk in tqdm(
                response.iter_content(chunk_size=1024),
                desc=desc,
                total=int(total_length / 1024) + 1,
            ):
                if chunk:
                    f.write(chunk)
                    f.flush()

    except Exception as e:
        print(f"Error when trying to download dataset @ {link}")
        raise e
    sys.stdout.write("\n")
    sys.stdout.flush()

__extract_zip(location, format)

Extracts the contents of a downloaded ZIP file and then deletes the zipped file.

Parameters:

Name Type Description Default
location str

filepath of the data directory that contains the ZIP file

required
format str

the format identifier string

required

Raises:

Type Description
RuntimeError

If there is an error unzipping the file

Source code in roboflow/core/version.py
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
def __extract_zip(self, location, format):
    """
    Extracts the contents of a downloaded ZIP file and then deletes the zipped file.

    Args:
        location (str): filepath of the data directory that contains the ZIP file
        format (str): the format identifier string

    Raises:
        RuntimeError: If there is an error unzipping the file
    """  # noqa: E501 // docs
    desc = None if TQDM_DISABLE else f"Extracting Dataset Version Zip to {location} in {format}:"
    with zipfile.ZipFile(location + "/roboflow.zip", "r") as zip_ref:
        for member in tqdm(
            zip_ref.infolist(),
            desc=desc,
        ):
            try:
                zip_ref.extract(member, location)
            except zipfile.error:
                raise RuntimeError("Error unzipping download")

    os.remove(location + "/roboflow.zip")

__get_download_location()

Get the local path to save a downloaded dataset to

Returns:

Name Type Description
str

the local path

Source code in roboflow/core/version.py
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
def __get_download_location(self):
    """
    Get the local path to save a downloaded dataset to

    Returns:
        str: the local path
    """
    version_slug = self.name.replace(" ", "-")
    filename = f"{version_slug}-{self.version}"

    directory = os.environ.get("DATASET_DIRECTORY")
    if directory:
        return f"{directory}/{filename}"

    return filename

__get_download_url(format)

Get the Roboflow API URL for downloading (and exporting downloadable zips)

Parameters:

Name Type Description Default
format str

the format identifier string

required

Returns:

Name Type Description
str

the Roboflow API URL

Source code in roboflow/core/version.py
611
612
613
614
615
616
617
618
619
620
621
622
def __get_download_url(self, format):
    """
    Get the Roboflow API URL for downloading (and exporting downloadable zips)

    Args:
        format (str): the format identifier string

    Returns:
        str: the Roboflow API URL
    """
    workspace, project, *_ = self.id.rsplit("/")
    return f"{API_URL}/{workspace}/{project}/{self.version}/{format}"

__get_format_identifier(format)

If format is none, fall back to the instance's model_format value.

If a human readable format name was passed, return the identifier that should be used for Roboflow API calls

Otherwise, assume that the passed in format is also the identifier

Parameters:

Name Type Description Default
format str

a human readable format string

required

Returns:

Name Type Description
str

format identifier string

Source code in roboflow/core/version.py
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
def __get_format_identifier(self, format):
    """
    If `format` is none, fall back to the instance's `model_format` value.

    If a human readable format name was passed, return the identifier that should be used for Roboflow API calls

    Otherwise, assume that the passed in format is also the identifier

    Args:
        format (str): a human readable format string

    Returns:
        str: format identifier string
    """  # noqa: E501 // docs
    if not format:
        format = self.model_format

    if not format:
        raise RuntimeError(
            "You must pass a format argument to version.download() or define a model in your Roboflow object"
        )

    friendly_formats = {"yolov5": "yolov5pytorch", "yolov7": "yolov7pytorch"}

    return friendly_formats.get(format, format)

__init__(version_dict, type, api_key, name, version, model_format, local, workspace, project, public, colors=None)

Initialize a Version object.

Source code in roboflow/core/version.py
 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
def __init__(
    self,
    version_dict,
    type,
    api_key,
    name,
    version,
    model_format,
    local: Optional[str],
    workspace,
    project,
    public,
    colors=None,
):
    """
    Initialize a Version object.
    """
    if api_key:
        self.__api_key = api_key
        self.name = name
        self.version = unwrap_version_id(version_id=version)
        self.type = type
        self.augmentation = version_dict["augmentation"]
        self.created = version_dict["created"]
        self.id = version_dict["id"]
        self.images = version_dict["images"]
        self.preprocessing = version_dict["preprocessing"]
        self.splits = version_dict["splits"]
        self.model_format = model_format
        self.workspace = workspace
        self.project = project
        self.public = public
        self.colors = {} if colors is None else colors

        self.colors = colors
        if "exports" in version_dict.keys():
            self.exports = version_dict["exports"]
        else:
            self.exports = []

        version_without_workspace = os.path.basename(str(version))

        response = requests.get(f"{API_URL}/{workspace}/{project}/{self.version}?api_key={self.__api_key}")
        if response.ok:
            version_info = response.json()["version"]
            has_model = bool(version_info.get("train", {}).get("model"))
        else:
            has_model = False

        if not has_model:
            self.model = None
        elif self.type == TYPE_OBJECT_DETECTION:
            self.model = ObjectDetectionModel(
                self.__api_key,
                self.id,
                self.name,
                version_without_workspace,
                local=local,
                colors=self.colors,
                preprocessing=self.preprocessing,
            )
        elif self.type == TYPE_CLASSICATION:
            self.model = ClassificationModel(
                self.__api_key,
                self.id,
                self.name,
                version_without_workspace,
                local=local,
                colors=self.colors,
                preprocessing=self.preprocessing,
            )
        elif self.type == TYPE_INSTANCE_SEGMENTATION:
            self.model = InstanceSegmentationModel(
                self.__api_key,
                self.id,
                colors=self.colors,
                preprocessing=self.preprocessing,
                local=local,
            )
        elif self.type == TYPE_SEMANTIC_SEGMENTATION:
            self.model = SemanticSegmentationModel(self.__api_key, self.id)
        elif self.type == TYPE_KEYPOINT_DETECTION:
            self.model = KeypointDetectionModel(self.__api_key, self.id, version=version_without_workspace)
        else:
            self.model = None

    elif DEMO_KEYS:
        api_key = DEMO_KEYS[0]
        if api_key == "coco-128-sample":
            self.__api_key = api_key
            self.model_format = model_format
            self.name = "coco-128"
            self.version = "1"
        else:
            self.__api_key = api_key
            self.model_format = model_format
            self.name = "chess-pieces-new"
            self.version = "23"
            self.id = "joseph-nelson/chess-pieces-new"

__reformat_yaml(location, format)

Certain formats seem to require reformatting the downloaded YAML.

Parameters:

Name Type Description Default
location str

filepath of the data directory that contains the yaml file

required
format str

the format identifier string

required
Source code in roboflow/core/version.py
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
def __reformat_yaml(self, location: str, format: str):
    """
    Certain formats seem to require reformatting the downloaded YAML.

    Args:
        location (str): filepath of the data directory that contains the yaml file
        format (str): the format identifier string
    """  # noqa: E501 // docs
    data_path = os.path.join(location, "data.yaml")

    def data_yaml_callback(content: dict) -> dict:
        if format == "mt-yolov6":
            content["train"] = location + content["train"].lstrip(".")
            content["val"] = location + content["val"].lstrip(".")
            content["test"] = location + content["test"].lstrip(".")
        if format in ["yolov5pytorch", "yolov7pytorch"]:
            content["train"] = location + content["train"].lstrip("..")
            content["val"] = location + content["val"].lstrip("..")
        try:
            # get_wrong_dependencies_versions raises exception if ultralytics is not installed at all  # noqa: E501 // docs
            if format == "yolov8" and not get_wrong_dependencies_versions(
                dependencies_versions=[("ultralytics", "==", "8.0.196")]
            ):
                content["train"] = "train/images"
                content["val"] = "valid/images"
                content["test"] = "test/images"
        except ModuleNotFoundError:
            pass
        return content

    if format in ["yolov5pytorch", "mt-yolov6", "yolov7pytorch", "yolov8", "yolov9"]:
        amend_data_yaml(path=data_path, callback=data_yaml_callback)

__str__()

String representation of version object.

Source code in roboflow/core/version.py
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
def __str__(self):
    """
    String representation of version object.
    """
    json_value = {
        "name": self.name,
        "type": self.type,
        "version": self.version,
        "augmentation": self.augmentation,
        "created": self.created,
        "preprocessing": self.preprocessing,
        "splits": self.splits,
        "workspace": self.workspace,
    }
    return json.dumps(json_value, indent=2)

deploy(model_type, model_path, filename='weights/best.pt')

Uploads provided weights file to Roboflow.

Parameters:

Name Type Description Default
model_type str

The type of the model to be deployed.

required
model_path str

File path to the model weights to be uploaded.

required
filename str

The name of the weights file. Defaults to "weights/best.pt".

'weights/best.pt'
Source code in roboflow/core/version.py
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
def deploy(self, model_type: str, model_path: str, filename: str = "weights/best.pt") -> None:
    """Uploads provided weights file to Roboflow.

    Args:
        model_type (str): The type of the model to be deployed.
        model_path (str): File path to the model weights to be uploaded.
        filename (str, optional): The name of the weights file. Defaults to "weights/best.pt".
    """
    model_type = normalize_yolo_model_type(model_type)
    zip_file_name = process(model_type, model_path, filename)

    if zip_file_name is None:
        raise RuntimeError("Failed to process model")

    self._upload_zip(model_type, model_path, zip_file_name)

download(model_format=None, location=None, overwrite=False)

Download and extract a ZIP of a version's dataset in a given format

:param model_format: A format to use for downloading :param location: An optional path for saving the file :param overwrite: An optional flag to prevent dataset overwrite when dataset is already downloaded

Parameters:

Name Type Description Default
model_format str

A format to use for downloading

None
location str

An optional path for saving the file

None
overwrite bool

An optional flag to overwrite an existing dataset if the dataset has already downloaded

False

Returns:

Type Description

Dataset Object

Raises:

Type Description
RuntimeError

If the Roboflow API returns an error with a helpful JSON body

HTTPError

If the Network/Roboflow API fails and does not return JSON

Source code in roboflow/core/version.py
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
def download(self, model_format=None, location=None, overwrite: bool = False):
    """
    Download and extract a ZIP of a version's dataset in a given format

    :param model_format: A format to use for downloading
    :param location: An optional path for saving the file
    :param overwrite: An optional flag to prevent dataset overwrite when dataset is already downloaded

    Args:
        model_format (str): A format to use for downloading
        location (str): An optional path for saving the file
        overwrite (bool): An optional flag to overwrite an existing dataset if the dataset has already downloaded

    Returns:
        Dataset Object

    Raises:
        RuntimeError: If the Roboflow API returns an error with a helpful JSON body
        HTTPError: If the Network/Roboflow API fails and does not return JSON
    """  # noqa: E501 // docs

    self.__wait_if_generating()

    model_format = self.__get_format_identifier(model_format)

    if model_format not in self.exports:
        self.export(model_format)

    # if model_format is not in

    if location is None:
        location = self.__get_download_location()
    if os.path.exists(location) and not overwrite:
        return Dataset(self.name, self.version, model_format, os.path.abspath(location))

    if self.__api_key == "coco-128-sample":
        link = "https://app.roboflow.com/ds/n9QwXwUK42?key=NnVCe2yMxP"
    else:
        url = self.__get_download_url(model_format)
        response = requests.get(url, params={"api_key": self.__api_key})
        if response.status_code == 200:
            link = response.json()["export"]["link"]
        else:
            try:
                raise RuntimeError(response.json())
            except json.JSONDecodeError:
                response.raise_for_status()

    self.__download_zip(link, location, model_format)
    self.__extract_zip(location, model_format)
    self.__reformat_yaml(location, model_format)  # TODO: is roboflow-python a place to be munging yaml files?

    return Dataset(self.name, self.version, model_format, os.path.abspath(location))

export(model_format=None)

Ask the Roboflow API to generate a version's dataset in a given format so that it can be downloaded via the download() method.

The export will be asynchronously generated and available for download after some amount of seconds - depending on dataset size.

Parameters:

Name Type Description Default
model_format str

A format to use for downloading

None

Returns:

Type Description

True

Raises:

Type Description
RuntimeError

If the Roboflow API returns an error with a helpful JSON body

HTTPError

If the Network/Roboflow API fails and does not return JSON

Source code in roboflow/core/version.py
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
def export(self, model_format=None):
    """
    Ask the Roboflow API to generate a version's dataset in a given format so that it can be downloaded via the `download()` method.

    The export will be asynchronously generated and available for download after some amount of seconds - depending on dataset size.

    Args:
        model_format (str): A format to use for downloading

    Returns:
        True

    Raises:
        RuntimeError: If the Roboflow API returns an error with a helpful JSON body
        HTTPError: If the Network/Roboflow API fails and does not return JSON
    """  # noqa: E501 // docs

    model_format = self.__get_format_identifier(model_format)

    self.__wait_if_generating()

    url = self.__get_download_url(model_format)
    response = requests.get(url, params={"api_key": self.__api_key})
    if not response.ok:
        try:
            raise RuntimeError(response.json())
        except json.JSONDecodeError:
            response.raise_for_status()

    # the rest api returns 202 if the export is still in progress
    if response.status_code == 202:
        status_code_check = 202
        while status_code_check == 202:
            time.sleep(1)
            response = requests.get(url, params={"api_key": self.__api_key})
            status_code_check = response.status_code
            if status_code_check == 202:
                progress = response.json()["progress"]
                progress_message = (
                    "Exporting format " + model_format + " in progress : " + str(round(progress * 100, 2)) + "%"
                )
                sys.stdout.write("\r" + progress_message)
                sys.stdout.flush()

    if response.status_code == 200:
        sys.stdout.write("\n")
        print("\r" + "Version export complete for " + model_format + " format")
        sys.stdout.flush()
        return True
    else:
        try:
            raise RuntimeError(response.json())
        except json.JSONDecodeError:
            response.raise_for_status()

train(speed=None, checkpoint=None, plot_in_notebook=False)

Ask the Roboflow API to train a previously exported version's dataset.

Parameters:

Name Type Description Default
speed

Whether to train quickly or accurately. Note: accurate training is a paid feature. Default speed is fast.

None
checkpoint

A string representing the checkpoint to use while training

None
plot

Whether to plot the training results. Default is False.

required

Returns:

Type Description
InferenceModel

An instance of the trained model class

Raises:

Type Description
RuntimeError

If the Roboflow API returns an error with a helpful JSON body

HTTPError

If the Network/Roboflow API fails and does not return JSON

Source code in roboflow/core/version.py
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
def train(self, speed=None, checkpoint=None, plot_in_notebook=False) -> InferenceModel:
    """
    Ask the Roboflow API to train a previously exported version's dataset.

    Args:
        speed: Whether to train quickly or accurately. Note: accurate training is a paid feature. Default speed is `fast`.
        checkpoint: A string representing the checkpoint to use while training
        plot: Whether to plot the training results. Default is `False`.

    Returns:
        An instance of the trained model class

    Raises:
        RuntimeError: If the Roboflow API returns an error with a helpful JSON body
        HTTPError: If the Network/Roboflow API fails and does not return JSON
    """  # noqa: E501 // docs

    self.__wait_if_generating()

    train_model_format = "yolov5pytorch"

    if self.type == TYPE_CLASSICATION:
        train_model_format = "folder"

    if self.type == TYPE_INSTANCE_SEGMENTATION:
        train_model_format = "yolov5pytorch"

    if self.type == TYPE_SEMANTIC_SEGMENTATION:
        train_model_format = "png-mask-semantic"

    # if classification
    if train_model_format not in self.exports:
        self.export(train_model_format)

    workspace, project, *_ = self.id.rsplit("/")
    url = f"{API_URL}/{workspace}/{project}/{self.version}/train"

    data = {}
    if speed:
        data["speed"] = speed

    if checkpoint:
        data["checkpoint"] = checkpoint

    write_line("Reaching out to Roboflow to start training...")

    response = requests.post(url, json=data, params={"api_key": self.__api_key})
    if not response.ok:
        try:
            raise RuntimeError(response.json())
        except json.JSONDecodeError:
            response.raise_for_status()

    status = "training"

    if plot_in_notebook:
        from IPython.display import clear_output
        from matplotlib import pyplot as plt

        def live_plot(epochs, mAP, loss, title=""):
            clear_output(wait=True)

            plt.subplot(2, 1, 1)
            plt.plot(epochs, mAP, "#00FFCE")
            plt.title(title)
            plt.ylabel("mAP")

            plt.subplot(2, 1, 2)
            plt.plot(epochs, loss, "#A351FB")
            plt.xlabel("epochs")
            plt.ylabel("loss")
            plt.show()

    first_graph_write = False
    previous_epochs: Union[np.ndarray, list] = []
    num_machine_spin_dots = []

    while status == "training" or status == "running":
        url = f"{API_URL}/{self.workspace}/{self.project}/{self.version}?nocache=true"
        response = requests.get(url, params={"api_key": self.__api_key})
        response.raise_for_status()
        version = response.json()["version"]
        if "models" in version.keys():
            models = version["models"]
        else:
            models = {}

        if "train" in version.keys():
            if "results" in version["train"].keys():
                status = "finished"
                break
            if "status" in version["train"].keys():
                if version["train"]["status"] == "failed":
                    write_line(line="Training failed")
                    break

        epochs: Union[np.ndarray, list]
        mAP: Union[np.ndarray, list]
        loss: Union[np.ndarray, list]

        if "roboflow-train" in models.keys():
            import numpy as np

            # training has started
            epochs = np.array([int(epoch["epoch"]) for epoch in models["roboflow-train"]["epochs"]])
            mAP = np.array([float(epoch["mAP"]) for epoch in models["roboflow-train"]["epochs"]])
            loss = np.array(
                [
                    sum(float(epoch[key]) for key in ["box_loss", "class_loss", "obj_loss"] if key in epoch)
                    for epoch in models["roboflow-train"]["epochs"]
                ]
            )

            title = "Training in Progress"
            # plottling logic
        else:
            num_machine_spin_dots.append(".")
            if len(num_machine_spin_dots) > 5:
                num_machine_spin_dots = ["."]
            title = "Training Machine Spinning Up" + "".join(num_machine_spin_dots)

            epochs = []
            mAP = []
            loss = []

        if (len(epochs) > len(previous_epochs)) or (len(epochs) == 0):
            if plot_in_notebook:
                live_plot(epochs, mAP, loss, title)
            else:
                if len(epochs) > 0:
                    title = (
                        title + ": Epoch: " + str(epochs[-1]) + " mAP: " + str(mAP[-1]) + " loss: " + str(loss[-1])
                    )
                if not first_graph_write:
                    write_line(title)
                    first_graph_write = True

        previous_epochs = copy.deepcopy(epochs)

        time.sleep(5)

    if not self.model:
        if self.type == TYPE_OBJECT_DETECTION:
            self.model = ObjectDetectionModel(
                self.__api_key,
                self.id,
                self.name,
                self.version,
                colors=self.colors,
                preprocessing=self.preprocessing,
            )
        elif self.type == TYPE_CLASSICATION:
            self.model = ClassificationModel(
                self.__api_key,
                self.id,
                self.name,
                self.version,
                colors=self.colors,
                preprocessing=self.preprocessing,
            )
        elif self.type == TYPE_INSTANCE_SEGMENTATION:
            self.model = InstanceSegmentationModel(
                self.__api_key,
                self.id,
                colors=self.colors,
                preprocessing=self.preprocessing,
            )
        elif self.type == TYPE_SEMANTIC_SEGMENTATION:
            self.model = SemanticSegmentationModel(self.__api_key, self.id)
        elif self.type == TYPE_KEYPOINT_DETECTION:
            self.model = KeypointDetectionModel(self.__api_key, self.id, version=self.version)
        else:
            raise ValueError(f"Unsupported model type: {self.type}")

    # return the model object
    assert self.model
    return self.model