The guide below describes creating a data exploration(EXPLORE) job. It uses PASCAL-VOC12 as an example public dataset. This public dataset can be imported as per the instructions in Import public dataset.
In the left navigation pane, click Repo > Datasets.
Select a dataset card and click Catalog.
For illustration purposes, let's consider creating an explore job for the “PASCAL-VOC12” dataset.On the catalog page, click the Visualize button in the top-right corner, as shown.
You can customize the table view as needed.
In the New Visualization Job window, enter the Job Name and Job Description.
Select the Job Type.
Enable the Region of Interest option to draw the bounding box in the image based on the area you are interested in for the job. Only the portion of the image within the bounding box will be considered to create the job. This is useful for cases where only a portion of the image has the entity that is of interest for analysis.
Notice that based on the bounding box you draw, the selected area is highlighted and the area outside the box is greyed out.
Enter other details, such as Clusterer, Embedder, and others, as needed.
Click Submit to create the visualization job.
Click the View Jobs link on the Notification, as shown.
Alternatively, navigate to Curate > Jobs.
The new job you just created will be listed as Created. Once the job creation is complete, the status changes to 'Ready'.
View Explore Jobs
Once you create an explore job, you can view it based on various actions.
Search jobs
On the top bar, enter the name or description of a job to search an explore job.
To search for a job based on the job type, select Explore from the Job Type drop down.
The Jobs page will display all the jobs of the type “Explore”.
To search for a job based on the status, select the required option from the Status drop down.
You can search jobs with the following statuses: Pending, Ready, Decayed, Failed, Submitted, Created, Processing.
To search for a job based on the filter options, click the Filter icon on the top-right corner, and select the parameters.
Dataset: Select the dataset attached to the job.
Clusterer: Select the clusterer name that was used for creating the job.
Tags: Select one or more tags if you had added tags while creating the job.
Pipeline: Select the required pipeline to filter by the featurizer that is used to create the job.
You can use one or more parameters in the Filter window to search for the jobs.
Explore Job Card Actions
In addition to the options for searching jobs, you can perform different activities on a specific job.
On the job card, click the 3-dots icon to view all the available options.
Clone: Select this option to clone an existing job. Enter the Name and Description for the job. The remaining options will be prefilled based on the job you have cloned from.
Archive: Select this option to archive the job. Once archived, the job will not appear by default on the page.
Enable the Show Archive option on the top-right corner to view the archived jobs.Unarchive: To unarchive an already archived job, from the 3-dots icon menu for that job card, select Unarchive.
The job gets unarchived and starts appearing on the job page by default.
Rename: Click this option to change the name of the job.
Edit tags: Click this option to modify the tags associated with the job. You can select a tag from the existing list or type a new tag and press Enter to add that tag.
Edit Description: Click this option to edit the job description, and then click Submit. The description overrides any existing description for that job.