Basic Search

TensorBay supports basic search based on different commits.

from tensorbay import GAS

# Please visit `https://gas.graviti.com/tensorbay/developer` to get the AccessKey.
gas = GAS("<YOUR_ACCESSKEY>")
dataset_client = gas.create_dataset("<DATASET_NAME>")

BasicSearchJob

TensorBay SDK allows create, get or list BasicSearchJob via BasicSearch.

Create

A BasicSearchJob can be created by create_job()

dataset_client.basic_search.create_job(
    title="search example",
    description="search description",
    conjunction="and",
    unit="file",
    filters=[
        (
            "category",
            "in",
            [
                "human.pedestrian.adult",
                "human.pedestrian.child",
                "human.pedestrian.construction_worker",
            ],
            "BOX3D",
        ),
        ("size", ">", 0),
        ("withLabel", "=", True),
        ("attribute", "in", {"attribute_1": [True, False], "attribute_2": [1, 2]}, "BOX3D"),
    ],
)

Note

conjunction: The logical conjunction between search filters, which includes “and” and “or”.

Note

unit:The unit of basic search. There are two options:

• “file”: Get the data that meets search filters;

• “frame”: If at least one data in a frame meets search filters, all data in the frame will be get. This option only works on fusion dataset.

Note

filters: The list of basic search criteria whose format is (key, operator, value, label_type).

• key: The key of filters, which could be “segment”, “size”, “withLabel”, “frame”, “sensor”, “category”, “attribute” or “keyword”. Here, the meaning of the “keyword” is the remote path of the data, such as “cat_01.jpg”.

• operator: The operational relationship between the key and value. The supported operators are “like”, “in”, “=”, “>”, “<”, “>=” and “<=”.

• value: The value of filters.

• label_type：It only needs to be used if the key is “category” or “attribute”, indicating the label type to which the category or attribute belongs.

There are also some restrictions on operators and values:

key	permitted operator	type of value
“segment”	“in”	list
“size”	“=”, “>”, “<”, “>=”, “<=”	int
“withLabel”	“=”	boolean
“frame”	“like”, “=”	string
“sensor”	“in”	list
“category”	“in”	list
“attribute”	“in”	dict
“keyword”	“like”, “=”	string

Get or List

The latest BasicSearchJob can be obtained by get_job() or list_jobs().

job = dataset_client.basic_search.get_job("<JOB_ID>")
job = dataset_client.basic_search.list_jobs()[0]

Get Information

Available BasicSearchJob information includes title, description, job_id, arguments, created_at, started_at, finished_at, status, error_message and result.

job.status
job.result
job.error_message
job.arguments

Note

If the BasicSearchJob is successfully completed, the result will be SearchResult or FusionSearchResult. See more details in Search Result.

Important

Only the latest five search results for one dataset can be used.

Update

The latest information of a BasicSearchJob can be obtained after update(). Note that if the until_complete is set to True, the BasicSearchJob will be blocked until it is completed.

job.update()
job.update(until_complete=True)

Abort

BasicSearchJob also supports abort():

job.abort()

Create Dataset

TensorBay SDK allows to create a dataset based on the search job by create_dataset()

job.create_dataset("<DATASET_NAME>")