Welcome to The Hasty Python API Library’s documentation!

Installation Guide

pip install hasty

Quick Start

Authentication

In order to communicate with your Hasty server via the API, you must provide valid API Key. The Key can be generated on Edit Workspace page (Hasty -> Edit Workspace -> API Accounts)

You will need to create a service account, define it’s role and finally generate API Key:

Create a Hasty API Client instance

This example shows how to create establish your initial connection to Hasty:

from hasty import Client

API_KEY = "bNZ09SA2hFSZGHa6jfMK2Ywo7GoActTXNCvJR1wEkVDtKvl9EMTLRvknwmwUz7Hjl1jPwxYBkyGF8BcWV3y9rg"

h = Client(api_key=API_KEY)

Managing projects

Below you can find an example of how the projects can be created, retrieved, updated or removed. Please note, that every project should belong to some workspace.

# Get workspaces that the user account can have access to
workspaces = h.get_workspaces()
print(workspaces)
>> ['Workspace(id="57d9a50e-73a2-47c5-a130-a652fa98d244", name="Hasty Python Library Workspace")']

# Create new project
default_workspace = workspaces[0]
new_project = h.create_project(workspace = default_workspace,
                           name = "My Awesome Project",
                           description = "Awesome description")
print(new_project)
>> Project(id="75472e11-d6f2-403c-80f6-f0fc83c97041", name="My Awesome Project")

# Get Projects
print(h.get_projects())
>> ['Project(id="75472e11-d6f2-403c-80f6-f0fc83c97041", name="My Awesome Project")']

# Get project by id
h.get_project(new_project.id)
>> Project(id="75472e11-d6f2-403c-80f6-f0fc83c97041", name="My Awesome Project")

# Edit project
new_project.edit(name="Super Awesome Project", description="Amazing description")

# Delete project
new_project.delete()

Managing images

Images in hasty stored in datasets. Similar to folders on your computer, every image should have a unique image name inside the dataset. You can upload image from local file or using url.

# Create dataset
train_dataset = new_project.create_dataset("train")
print(train_dataset)
>> Dataset(id="7a55886d-e695-4693-9a3d-7addd75c5e74", name="train")

# Upload image from file
image = new_project.upload_from_file(dataset=train_dataset,
                                     filepath='../Datasets/African Wildlife/rhino/001.jpg')
print(image)
>> Image(id="b58860f1-8da0-4c94-bff7-b7476c3c8f50", dataset_name="train", name="001.jpg")

# Upload from URL
image = new_project.upload_from_url(dataset=train_dataset,
                                    filename='4.jpg',
                                    url='https://www.gstatic.com/webp/gallery/4.jpg')
print(image)
>> Image(id="2dae09ce-ca8a-416f-90e2-a4024515ae95", dataset_name=None, name="4.jpg")

# Retrieve the list of projects images
images = new_project.get_images()
print(images)
>> ['Image(id="b58860f1-8da0-4c94-bff7-b7476c3c8f50", dataset_name="train", name="001.jpg")',
    'Image(id="2dae09ce-ca8a-416f-90e2-a4024515ae95", dataset_name="train", name="4.jpg")']

Managing label classes

Every label in hasty should belongs to some label class.

# Create label classes
rhino_class = new_project.create_label_class(name="rhino", color="#6a3d9a", class_type="object")
sky_class = new_project.create_label_class(name="sky", color="#6a3d9a", class_type="background")

# Edit label class
sky_class.edit(name="sky", color="#f0e928", class_type="background")

# Get label classes
label_classes = new_project.get_label_classes()
print(label_classes)
>> ['LabelClass(id="1201c994-c0dc-4efa-9892-f9d030320c5d", name="rhino", color="#6a3d9a", type="object", norder=10.0)',
    'LabelClass(id="2c8097bc-6dbf-4753-9485-683aff6171f9", name="sky", color="#f0e928", type="background", norder=11.0)']

# Delete label class
sky_class.delete()

Managing labels

# Create label
image.create_label(label_class=rhino_class, bbox=[20, 30, 300, 400])
lbl.edit(label_class=rhino_class, bbox=[120, 130, 300, 400])
lbl.delete()

Python API

`Client`

class hasty.Client(api_key, base_url='https://api.hasty.ai')

Client for Hasty API

__init__(api_key, base_url='https://api.hasty.ai')

Initialize the client

api_key: Your API key

get_workspaces(): Returns the list of workspaces that the user can have an access to :returns: A list of Workspace objects.

get_projects(): Returns the list of projects :returns: A list of Project objects.

get_project(project_id)

Returns project Project by id

Parameters: project_id (str) – Project id

create_project(workspace: Union[str, Workspace], name: str, description: Optional[str] = None) → Project

Creates new project Project

Parameters

workspace (Workspace, str) – Workspace object or id which the project should belongs to
name (str) – Name of the project
description (str, optional) – Project description

`Project`

class hasty.Project(requester, data, obj_params=None)

property id: type: string

property name: type: string

property description: type: string

property workspace_id: type: string

edit(name, description)

Edits projects properties

Parameters

name (str) – Name of the project
description (str, optional) – Project description

delete(): Removes project

get_datasets(): Returns projects datasets Dataset objects.

get_dataset(dataset_id: str)

Get dataset by id, returns ~hasty.Dataset object

Parameters: dataset_id (str) – Dataset id

create_dataset(name: str, norder: float = 0)

Creates a new dataset, returns ~hasty.Dataset object

Parameters

name (str) – Name of the dataset
norder (float, optional) – Order in the list

get_images(dataset=None, image_status=None)

Retrieves the list of projects images.

Parameters

dataset (str, ~hasty.Dataset, list of str, list of ~hasty.Dataset) – filter images by dataset
image_status (str, list of str) –
Filters images by status, valid values are:
- ”NEW”
- ”DONE”,
- ”SKIPPED”
- ”IN PROGRESS”
- ”TO REVIEW”
- ”AUTO-LABELLED”

get_image(image_id)

Retrieves the image by its id.

Parameters: image_id (str) – Image ID

upload_from_file(dataset, filepath, external_id: Optional[str] = None)

Uploads image from the given filepath

Parameters

dataset (~hasty.Dataset, str) – Dataset object or id that the image should belongs to
filepath (str) – Local path
external_id (str) – External ID (optional)

upload_from_url(dataset: Union[Dataset, str], filename: str, url: str, copy_original: bool = True, external_id: Optional[str] = None)

Uploads image from a given URL

Parameters

dataset (~hasty.Dataset, str) – Dataset object or id that the image should belongs to
filename (str) – Filename of the image
url (str) – Image url
copy_original (str) – If True Hasty makes a copy of the image. Default True.
external_id (str) – External ID (optional)

get_label_classes(): Get label classes, list of LabelClass objects.

get_label_class(label_class_id: str)

Get label class by id, returns ~hasty.LabelClass object

Parameters: label_class_id (str) – Label class id

create_label_class(name: str, color: Optional[str] = None, class_type: str = 'object', norder: Optional[float] = None, external_id: Optional[str] = None)

Create label class, returns LabelClass object.

Parameters

name (str) – Label class name
color (str, optional) – Color in HEX format #0f0f0faa
class_type (str, optional) – Class type [object or background] (default object)
norder (float, optional) – Order in the Hasty tool
external_id (str, optional) – External Identifier

get_tag_classes(): Get tag classes, list of TagClass objects.

get_tag_class(tag_class_id: str)

Get tag class by id, returns ~hasty.TagClass object

Parameters: tag_class_id (str) – Tag class id

create_tag_class(name: str, norder: Optional[float] = None)

Create tag class, returns TagClass object.

Parameters

name (str) – Tag class name
norder (float, optional) – Order in the Hasty tool

get_attributes(): Get label classes, list of Attribute objects.

create_attribute(name: str, attribute_type: str, description: Optional[str] = None, norder: Optional[float] = None, values: Optional[List[str]] = None)

Create attribute, returns Attribute object.

Parameters

name (str) – Attribute name
attribute_type (str) – Attribute type [‘SELECTION’, ‘MULTIPLE-SELECTION’, ‘TEXT’, ‘INT’, ‘FLOAT’, ‘BOOL’]
description (str, optional) – Attrbute description
norder (float, optional) – Order in the Hasty tool
values (list of str) – List of values for SELECTION and MULTIPLE-SELECTION attribute type

get_attribute_classes()

Get attributes - class mapping. Returns list of dict with a keys:

attribute_id - Attribute ID
class_id - Class ID
attribute_order - Order of attributes within the class
class_order - Order of classes within the attribute

set_attribute_classes(attribute_classes)

Set attribute - class mapping

Parameters

attribute_classes (dict) –
ID (- class_id - Class) –
ID –
class (- attribute_order - Order of attributes within the) –
attribute (- class_order - Order of classes within the) –

delete_attribute_classes(attribute_classes)

Removes attribute - class mapping

Parameters

attribute_classes (dict) –
ID (- class_id - Class) –
ID –
class (- attribute_order - Order of attributes within the) –
attribute (- class_order - Order of classes within the) –

export(name: str, export_format: str, dataset: Optional[Union[Dataset, str, List[Dataset], List[str]]] = None, image_status: Union[str, List[str]] = 'DONE', sign_urls: bool = False, semantic_format: str = 'gs_asc', labels_order: Optional[List[str]] = None)

Initiate export job. Returns ExportJob object.

Parameters

name (str) – Name of the export file
export_format (str) – Export format one of [“json_v1.1”, “semantic_png”, “json_coco”, “images”]
dataset (~hasty.Dataset, str, list of ~hasty.Dataset, list of str) – List of the datasets to export
image_status (list of str, str) – List of the image statuses to export. Default DONE
sign_urls (bool) – Whether to generate sign urls for images. Default False
semantic_format (str) – Format for semantic_png export. [“gs_desc”, “gs_asc”, “class_color”]
labels_order (list of str) – Draw order for semantic_png export [“z_index”, “class_type”, “class_order”]

get_detector(): Returns object detection model. Returns Detector object.

get_instance_segmentor(): Returns instance segmentation model. Returns InstanceSegmentor object.

get_semantic_segmentor(): Returns semantic segmentation model. Returns SemanticSegmentor object.

get_attributer(): Returns active attributer model for the project. Returns Attributer object.

get_automated_labeling_jobs(): Get automated labeling jobs, list of AutomatedLabelingJob objects.

get_automated_labeling_job(job_id: str)

Get automated labeling jobs, list of AutomatedLabelingJob objects.

Parameters: job_id (str) – Automated labeling job id

create_automated_labeling_job(experiment_id: str, confidence_threshold: float = 0.8, max_detections_per_image: int = 100, num_images: int = 0, masker_threshold: float = 0.5, dataset_id: Optional[str] = None)

Create automated labeling job, returns AutomatedLabelingJob object.

Parameters

experiment_id (str) – ID of an experiment, that would be used for automated labeling
confidence_threshold (float, optional) – Confidence threshold of the predictions that should be applied (default 0.8)
max_detections_per_image (int, optional) – Max number of labels that should be created, default 100
num_images (int, optional) – Total number of images that should be used for automated labeling (default all)
masker_threshold (float, optional) – Threshold for mask head (default 0.5)
dataset_id (str, optional) – Filter images by dataset

`Dataset`

class hasty.Dataset(requester, data, obj_params=None)

property id: type: string

property name: type: string

property project_id: type: string

property norder: type: float

edit(name: str, norder: float)

Edit dataset properties

Parameters

name (str) – Name of the dataset
norder (float) – Order in the list

delete(): Removes dataset

`Image`

class hasty.Image(requester, data, obj_params=None)

property id: type: string

property name: type: string

property project_id: type: string

property dataset_id: type: string

property dataset_name: type: string

property width: type: int

property height: type: int

property status: type: string

property public_url: type: string

property external_id: type: string

get_labels(): Returns image labels (list of ~hasty.Label objects)

create_label(label_class: Union[LabelClass, str], bbox: Optional[List[int]] = None, polygon: Optional[List[List[int]]] = None, mask: Optional[List[int]] = None, z_index: Optional[int] = None, external_id: Optional[str] = None)

Create label

Parameters

label_class (LabelClass, str) – Label class or label class ID of the label
bbox (list of int) – Coordinates of bounding box [x_min, y_min, x_max, y_max]
polygon (list) – List of x, y pairs [[x0, y0], [x1, y1], …. [x0, y0]]
mask (list of int) – RLE Encoded binary mask, (order right -> down)
z_index (int) – Z index of the label. A label with greater value is in front of a label with a lower one.
external_id (str, optional) – External Identifier

create_labels(labels)

Create multiple labels. Returns a list of ~hasty.Label objects

Parameters: labels (list of dict) – List of labels, keys: class_id: Label class ID of the label bbox: Coordinates of bounding box [x_min, y_min, x_max, y_max] polygon: List of x, y pairs [[x0, y0], [x1, y1], …. [x0, y0]] mask: RLE Encoded binary mask, (order right -> down) z_index: Z index of the label. external_id: External Identifier

edit_labels(labels)

Updates multiple labels. Returns a list of ~hasty.Label objects

Parameters: labels (list of dict) – List of labels, keys: label_id: Label id class_id: Label class ID of the label bbox: Coordinates of bounding box [x_min, y_min, x_max, y_max] polygon: List of x, y pairs [[x0, y0], [x1, y1], …. [x0, y0]] mask: RLE Encoded binary mask, (order right -> down) z_index: Z index of the label. external_id: External identifier

delete_labels(label_ids: List[str])

Removes multiple labels

Parameters: label_ids (list of str) – Returns list of ids

set_status(status: str)

Set image status

Parameters: status – New status one of [“NEW”, “DONE”, “SKIPPED”, “IN PROGRESS”, “TO REVIEW”]

download(filepath: str)

Downloads image to file

Parameters: filepath (str) – Local path

rename(new_name: str)

Rename image

Parameters: new_name (str) – New image name

move(dataset: Union[Dataset, str]): Move image to another dataset

delete(): Removes image

get_tags(): Returns image tags (list of ~hasty.Tag objects)

add_tags(tags: List[Union[Dict, TagClass]])

Create multiple tags. Returns a list of ~hasty.Tag objects

Parameters: tags (list of dict/~hasty.TagClass) – List of tags, keys: tag_class_id: Tag class ID

delete_tags(tags: List[Union[Dict, Tag, TagClass]])

Removes multiple tags

Parameters: tags (list of dict/~hasty.Tag/~hasty.TagClass) – List of tags, keys: tag_id: Tag ID of the label (optional if tag_class_id is specified) tag_class_id: Tag class ID of the label (optional if id is specified)

`LabelClass`

class hasty.LabelClass(requester, data, obj_params=None)

property id: type: string

property name: type: string

property project_id: type: string

property color: type: string

property class_type: type: string

property norder: type: float

property external_id: type: string

edit(name, color=None, class_type='object', norder=None, external_id=None)

Edit label class properties

Parameters

name (str) – Label class name
color (str, optional) – Color in HEX format #0f0f0faa
class_type (str, optional) – Class type [object or background] (default object)
norder (float, optional) – Order in the Hasty tool
external_id (str, optional) – External identifier

delete(): Deletes label class

`TagClass`

class hasty.TagClass(requester, data, obj_params=None)

property id: type: string

property name: type: string

property project_id: type: string

property norder: type: float

edit(name, norder=None)

Edit tag class properties

Parameters

name (str) – Taп class name
norder (float, optional) – Order in the Hasty tool

delete(): Deletes tag class

`Attribute`

class hasty.Attribute(requester, data, obj_params=None)

property id: type: string

property name: type: string

property project_id: type: string

property attribute_type: type: string

property description: type: string

property norder: type: float

property values: type: Dict

edit(name: str, attribute_type: str, description: Optional[str] = None, norder: Optional[float] = None, values: Optional[List[str]] = None)

Edit attribute properties

Parameters

Args –
name (str) – Attribute name
attribute_type (str) – Attribute type [‘SELECTION’, ‘MULTIPLE-SELECTION’, ‘TEXT’, ‘INT’, ‘FLOAT’, ‘BOOL’]
description (str, optional) – Attrbute description
norder (float, optional) – Order in the Hasty tool
values (list of str) – List of values for SELECTION and MULTIPLE-SELECTION attribute type

delete(): Deletes attribute

`Label`

class hasty.Label(requester, data, obj_params=None)

property id: type: string

property project_id: type: string

property image_id: type: string

property class_id: type: string

property bbox: type: string

property polygon: type: string

property mask: type: string

property z_index: type: string

property external_id: type: string

edit(label_class, bbox=None, polygon=None, mask=None, z_index=None, external_id=None)

Update label properties

Parameters

label_class (LabelClass, str) – Label class or label class ID of the label
bbox (list of int) – Coordinates of bounding box [x_min, y_min, x_max, y_max]
polygon (list) – List of x, y pairs [[x0, y0], [x1, y1], …. [x0, y0]]
mask (list of int) – RLE Encoded binary mask, (order right -> down)
z_index (float) – Z index of the label. A label with greater value is in front of a label with a lower one.
external_id (str, optional) – External Identifier

delete(): Delete label

get_attributes(): Get attributes values, list of LabelAttribute objects.

set_attribute(attribute, value)

Set label attribute

Parameters

attribute (~hasty.Attribute, str) –
value (str, float, int, bool, list of str) –

`Tag`

class hasty.Tag(requester, data, obj_params=None)

property id: type: string

property tag_class_id: type: string

property project_id: type: string

property image_id: type: string

property tag_class_name: type: string

`Attributer`

class hasty.Attributer(requester, data, obj_params=None)

discover_model(): Performs model discovery and loads model to GPU

predict(image_url: Optional[str] = None, image_path: Optional[str] = None, bboxes: Optional[List[List[str]]] = None, confidence_threshold: float = 0.5)

Returns predictions for provided image.

Parameters

image_url (str) – Image URL
image_path (str) – Path to local image file
bboxes (list of list of int) – List of bounding boxes [x_min, y_min, x_max, y_max]
confidence_threshold (float) – Confidence threshold [0, 1) (default 0.5)

Returns

bbox (list of int): Coordinates of bounding box [x_min, y_min, x_max, y_max]
attribute_id (str): Attribute id
lov_ids (list of str): List of attribute values

Return type

List of List of dict

`Detector`

class hasty.Detector(requester, data, obj_params=None)

discover_model(): Performs model discovery and loads model to GPU

predict(image_url: Optional[str] = None, image_path: Optional[str] = None, confidence_threshold: float = 0.5, max_detections_per_image: int = 100)

Returns predictions for provided image.

Parameters

image_url (str) – Image URL
image_path (str) – Path to local image file
confidence_threshold (float) – Confidence threshold [0, 1) (default 0.5)
max_detections_per_image (int) – Maximum detections per image (default 100)

Returns

bbox (list of int): Coordinates of bounding box [x_min, y_min, x_max, y_max]
score (float): Confidence score
class_id (str): Label class ID

Return type

List of dict

`InstanceSegmentor`

class hasty.InstanceSegmentor(requester, data, obj_params=None)

discover_model(): Performs model discovery and loads model to GPU

predict(image_url: Optional[str] = None, image_path: Optional[str] = None, confidence_threshold: float = 0.5, max_detections_per_image: int = 100)

Returns predictions for provided image.

Parameters

image_url (str) – Image URL
image_path (str) – Path to local image file
confidence_threshold (float) – Confidence threshold [0, 1) (default 0.5)
max_detections_per_image (int) – Maximum detections per image (default 100)

Returns

bbox (list of int): Coordinates of bounding box [x_min, y_min, x_max, y_max]
mask (list of int): RLE Encoded binary mask, (order right -> down)
score (float): Confidence score
class_id (str): Label class ID

Return type

List of dict

`SemanticSegmentor`

class hasty.SemanticSegmentor(requester, data, obj_params=None)

discover_model(): Performs model discovery and loads model to GPU

predict(image_url: Optional[str] = None, image_path: Optional[str] = None, min_size: float = 0.0)

Returns predictions for provided image.

Parameters

image_url (str) – Image URL
image_path (str) – Path to local image file
min_size (float) – Ignore prediction with an area less then min_size

Returns

bbox (list of int): Coordinates of bounding box [x_min, y_min, x_max, y_max]
mask (list of int): RLE Encoded binary mask, (order right -> down)
class_id (str): Label class ID

Return type

List of dict

Label Utils

hasty.label_utils.rle_encoding(x: array)

Encode binary mask to RLE

Parameters: x (np.array) – numpy array of shape (height, width), 1 - mask, 0 - background

Returns run length as list

hasty.label_utils.rle_decode(mask_rle: List[int], shape: Tuple[int])

Decodes mask from RLE format to binary mask. Returns numpy array, 1 - mask, 0 - background

Parameters

mask_rle (list) – run-length as string formated (start length)
shape (tuple) – (width, height) of array to return

hasty.label_utils.yolo2hasty(yolo_bbox: List[float], image_width: int, image_height: int)

Converts normalized YOLO bbox format (Xn_center, Yn_center, Wn, Hn) to Hasty format (x_min, y_min, x_max, y_max)

Parameters

yolo_bbox (list of floats) – Normalized coordinates [X_center, Y_center, W, H] example [0.3, 0.4, 0.1, 0.25]
image_width (int) – Image width
image_height (int) – Image height

hasty.label_utils.polygon2box(polygon: List[List[int]])

Converts polygon to bounding box [x_min, y_min, x_max, y_max]

Parameters: polygon – List of x, y pairs [[x0, y0], [x1, y1], …. [x0, y0]]

Welcome to The Hasty Python API Library’s documentation!

Installation Guide

Quick Start

Authentication

Create a Hasty API Client instance

Managing projects

Managing images

Managing label classes

Managing labels

Python API

Label Utils

Indices and tables