Dataset¶

class Dataset[source]¶

Bases: supervisely.collection.key_indexed_collection.KeyObject

Dataset is where your labeled and unlabeled images and other data files live. Dataset object is immutable.

Parameters

directory : str: Path to dataset directory.
mode : OpenMode, optional. If not provided, dataset_id must be provided.: Determines working mode for the given dataset.
parents : List[str]: List of parent directories, e.g. [“ds1”, “ds2”, “ds3”].
dataset_id : Optional[int]: Dataset ID if the Dataset is opened in API mode. If dataset_id is specified then api must be specified as well.
api : Optional[:class:`Api<supervis: API object if the Dataset is opened in API mode.

Usage example

import supervisely as sly
dataset_path = "/home/admin/work/supervisely/projects/lemons_annotated/ds1"

# To open dataset locally in read mode
ds = sly.Dataset(dataset_path, sly.OpenMode.READ)

# To open dataset on API
api = sly.Api.from_env()
ds = sly.Dataset(dataset_path, dataset_id=1, api=api)

Methods

`add_item_file`	Adds given item file to dataset items directory, and adds given annotation to dataset annotations directory.
`add_item_np`	Adds given numpy matrix as an image to dataset items directory, and adds given annotation to dataset ann directory.
`add_item_raw_bytes`	Adds given binary object as an image to dataset items directory, and adds given annotation to dataset ann directory.
`datasets_dir`	rtype `List`[`str`]
`delete_item`	Delete image, image info and annotation from `Dataset`.
`generate_item_path`	Generates full path to the given item.
`get_ann`	Read annotation of item from json.
`get_ann_path`	Path to the given annotation json file.
`get_classes_stats`
`get_image_info`	Information for Item with given name.
`get_img_info_path`	Get path to the image info json file without checking if the file exists.
`get_img_path`	Path to the given image.
`get_item_info`	Information for Item with given name.
`get_item_info_path`	Get path to the item info json file without checking if the file exists.
`get_item_meta_path`	Get path to the item info json file without checking if the file exists.
`get_item_path`	Path to the given item.
`get_item_paths`	Generates `ItemPaths` object with paths to item and annotation directories for item with given name.
`get_items_names`	List of dataset item names.
`get_seg_path`	Get path to the png segmentation mask file without checking if the file exists.
`get_url`	Get URL to dataset items list in Supervisely.
`ignorable_dirs`	rtype `List`[`str`]
`item_exists`	Checks if given item name belongs to the dataset.
`items`	This method is used to iterate over dataset items, receiving item name, path to image and path to annotation json file.
`key`
`set_ann`	Replaces given annotation for given item name to dataset annotations directory in json format.
`set_ann_dict`	Replaces given annotation json for given item name to dataset annotations directory in json format.
`set_ann_file`	Replaces given annotation json file for given item name to dataset annotations directory in json format.

Attributes

`ann_dir`	Path to the dataset annotations directory.
`ann_dir_name`
`datasets_dir_name`
`directory`	Path to the dataset directory.
`image_infos`	If the dataset is opened from the API, returns the list of ImageInfo objects.
`img_dir`	Path to the dataset images directory.
`img_info_dir`	Path to the dataset image info directory.
`item_dir`	Path to the dataset items directory.
`item_dir_name`
`item_info_dir`	Path to the dataset item info directory.
`item_info_dir_name`
`meta_dir`	Path to the dataset segmentation masks directory.
`meta_dir_name`
`name`	Full Dataset name, which includes it's parents, e.g.
`path`	Returns a relative local path to the dataset.
`project_dir`	Path to the project containing the dataset.
`seg_dir`	Path to the dataset segmentation masks directory.
`seg_dir_name`
`short_name`	Short dataset name, which does not include it's parents.

annotation_class¶: alias of supervisely.annotation.annotation.Annotation

item_info_class¶: alias of supervisely.api.image_api.ImageInfo

add_item_file(item_name, item_path, ann=None, _validate_item=True, _use_hardlink=False, item_info=None, img_info=None)[source]¶

Adds given item file to dataset items directory, and adds given annotation to dataset annotations directory. if ann is None, creates empty annotation file.

Parameters

item_name : str: Item name.
item_path : str: Path to the item.
ann : Annotation or str, optional: Annotation object or path to annotation json file.
_validate_item : bool, optional: Checks input files format.
_use_hardlink : bool, optional: If True creates a hardlink pointing to src named dst, otherwise don’t.
item_info : ImageInfo or dict or str, optional: ImageInfo object or ImageInfo object converted to dict or path to item info json file for copying to dataset item info directory.
img_info : ImageInfo or dict or str, optional: Deprecated version of item_info parameter. Can be removed in future versions.

Returns

None

Return type

NoneType

Raises

RuntimeError if item_name already exists in dataset or item name has unsupported extension.

Usage example

import supervisely as sly
dataset_path = "/home/admin/work/supervisely/projects/lemons_annotated/ds1"
ds = sly.Dataset(dataset_path, sly.OpenMode.READ)

ann = "/home/admin/work/supervisely/projects/lemons_annotated/ds1/ann/IMG_8888.jpeg.json"
ds.add_item_file("IMG_8888.jpeg", "/home/admin/work/supervisely/projects/lemons_annotated/ds1/img/IMG_8888.jpeg", ann=ann)
print(ds.item_exists("IMG_8888.jpeg"))
# Output: True

add_item_np(item_name, img, ann=None, img_info=None)[source]¶

Adds given numpy matrix as an image to dataset items directory, and adds given annotation to dataset ann directory. if ann is None, creates empty annotation file.

Parameters

item_name : str: Item name.
img : np.ndarray: numpy Image matrix in RGB format.
ann : Annotation or str, optional: Annotation object or path to annotation json file.
img_info : ImageInfo or dict or str, optional: ImageInfo object or ImageInfo object converted to dict or path to item info json file for copying to dataset item info directory.

Returns

None

Return type

NoneType

Raises

RuntimeError if item_name already exists in dataset or item name has unsupported extension

Usage example

import supervisely as sly
dataset_path = "/home/admin/work/supervisely/projects/lemons_annotated/ds1"
ds = sly.Dataset(dataset_path, sly.OpenMode.READ)

img_path = "/home/admin/Pictures/Clouds.jpeg"
img_np = sly.image.read(img_path)
ds.add_item_np("IMG_050.jpeg", img_np)
print(ds.item_exists("IMG_050.jpeg"))
# Output: True

add_item_raw_bytes(item_name, item_raw_bytes, ann=None, img_info=None)[source]¶

Adds given binary object as an image to dataset items directory, and adds given annotation to dataset ann directory. if ann is None, creates empty annotation file.

Parameters

item_name : str: Item name.
item_raw_bytes : bytes: Binary object.
ann : Annotation or str, optional: Annotation object or path to annotation json file.
img_info : ImageInfo or dict or str, optional: ImageInfo object or ImageInfo object converted to dict or path to item info json file for copying to dataset item info directory.

Returns

None

Return type

NoneType

Raises

RuntimeError if item_name already exists in dataset or item name has unsupported extension

Usage example

import supervisely as sly
dataset_path = "/home/admin/work/supervisely/projects/lemons_annotated/ds1"
ds = sly.Dataset(dataset_path, sly.OpenMode.READ)

img_path = "/home/admin/Pictures/Clouds.jpeg"
img_np = sly.image.read(img_path)
img_bytes = sly.image.write_bytes(img_np, "jpeg")
ds.add_item_raw_bytes("IMG_050.jpeg", img_bytes)
print(ds.item_exists("IMG_050.jpeg"))
# Output: True

delete_item(item_name)[source]¶

Delete image, image info and annotation from Dataset.

Parameters

item_name : str: Item name.

Returns

True if item was successfully deleted, False if item wasn’t found in dataset.

Return type

bool

Usage example

import supervisely as sly
dataset_path = "/home/admin/work/supervisely/projects/lemons_annotated/ds1"
ds = sly.Dataset(dataset_path, sly.OpenMode.READ)

print(dataset.delete_item("IMG_0748"))
# Output: False

print(dataset.delete_item("IMG_0748.jpeg"))
# Output: True

generate_item_path(item_name)[source]¶

Generates full path to the given item.

Parameters

item_name : str: Item name.

Returns

Full path to the given item

Return type