Create new dataset
Last updated
Last updated
Create a new dataset in a client.
** From 1st January 2020, one needs to call after this API call to complete preprocessing.
URL
POST
Request
Path parameters
Parameter name
Parameter type
Description
clientId
string
Client Id obtained from Deepen AI
Request body
Property name
Property type
Description
files
array
Array of metadata objects, each representing a file to be uploaded. Please see the example object below for the structure of the metadata object. For datasets of dataset_type ‘images’’, the metadata should be of a zip file of all images to be uploaded as part of the dataset. For datasets of dataset_type ‘video’, the metadata should be of an mp4 of the video to be uploaded as part of the dataset.
dataset_format
string
The format in which the dataset is uploaded for ‘3d’ dataset types. Can take one of the following values :
json
default
Not required for dataset_type other than ‘3d’
dataset_type
string
The type of uploaded dataset:
3d
video
images
dataset_name
string
Name of the uploaded dataset
labelling_mode
string
Labelling format in the dataset:
label_by_label (For sequence datasets)
frame_by_frame (For single frame datasets - default)
labelling_profile_id
string
Labelling profile id
This is the profile id of the profile from which to get the labelling instructions.
labelling_profile
object
LabellingProfileResource indicating the labelling instructions. Either labelling_profile or labelling_profile_id is necessary to include labelling instructions with the dataset.
tags
array
Array of TagResource
user_scopes
array
Array of DatasetUserScopeResource
user_group_scopes
array
Array of DatasetUserGroupScopeResource
Example body
{ 'dataset_name': 'testDataset',
'dataset_type': '3d',
'fps': 10,
'labelling_profile': {'pipeline_spec': [{'stage_id': 'Labelling', 'can_edit': True, 'skills': {'box': {}, 'point': {}, 'polygon': {}, 'lane': {}, '3d_bbox': {}, '3d_point': {}, '3d_instance_point': {}, '3d_polygon': {}, '3d_polyline': {}}}, {'stage_id': 'QA', 'can_edit': False, 'skills': {'box': {}, 'point': {}, 'polygon': {}, 'lane': {}, '3d_bbox': {}, '3d_point': {}, '3d_instance_point': {}, '3d_polygon': {}, '3d_polyline': {}}}], 'features': {'box': {'attributes': {}, 'label_category_ids': []}, 'lane': {'attributes': {}, 'label_category_ids': []}, 'polygon': {'attributes': {}, 'label_category_ids': []}, 'point': {'attributes': {}, 'label_category_ids': []}, '3d_bbox': {'attributes': {}, 'label_category_ids': []}, '3d_polyline': {'attributes': {}, 'label_category_ids': []}, '3d_polygon': {'attributes': {}, 'label_category_ids': []}, '3d_point': {'attributes': {}, 'label_category_ids': []}, '3d_instance_point': {'attributes': {}, 'label_category_ids': []}}, 'instructions': {'text': ''}, 'notification_url': ''},
'labelling_mode': 'frame_by_frame',
'dataset_format': 'default',
'sensor_maker': None,
'files': [{'file_size': 16513471, 'file_type': 'application/zip'}] }
Response
Returns an array of resumable upload urls on successful completion
Response body
Property name
Property type
Description
files
array
dataset_id
string
The new dataset id created for the uploaded dataset
Example
{
'dataset_id': 'BBZztvi0Z9TC',
'files': [{'resumable_upload_url': 'https://www.googleapis.com/upload/storage/v2/b/deepenai-dev-rannotateclientdata/o?uploadType=resumable&name=rannotate_client_uploaded_data/8cz4s0r1/defaultproject/BBZztvi0Z9TC/3d_data.zip&upload_id=AAANsUnQh_s03FsS-iGuvfD9n3ABpmvLxV8-Yi86g-QwGrtntj3HfYE8e_l1iDbeKAp06MXNGqvbotpgti1n3yjQEF0', 'signed_form_data': None, 'file_size': 16513471, 'file_type': 'application/zip'}] }
Please use the below command or follow this link to upload the data
curl -v -X PUT -H 'Expect:' --upload-file '<path to zip file>' '<resumable upload url>'
Array of metadata objects, each representing a file to be uploaded. Please see the example object below for the structure of the metadata object. Each metadata object has a Google cloud storage resumable upload url, to which the file has to be uploaded. Please see google cloud storage documentation at to learn how to upload the file using the url.
Once the dataset zip file is uploaded, you need to call the to process the zip file and make it ready for labelling.