Tagging system

Organizing your Data with Picsellia's DataTag.

In your Organization's Datalake, where all your Data is stored, things can become disorganized after weeks of work – often referred to as "data chaos."

The primary goal of the tools within the Datalake is to assist you in structuring and organizing your Data as needed.

1. Introducing DataTag

Think of a tag as a marker that you can name and connect with different elements in Picsellia. Its primary purpose is to make organizing and locating these elements easier. Since almost everything in Picsellia can have a tag – whether it's a Datalake, a Dataset, and more – the ones linked to the Datalake are called DataTag.

For this page, we're only focusing on the DataTag.

Remember, each Data can have multiple DataTag.

This additional layer of organization empowers you to create a unique hierarchy for your Data within the Datalake. Even though all your Data is stored in one place, you can still find and navigate through them based on specific DataTag.

Of course, you can use DataTag in any way that suits your needs, allowing you to build your Datalake structure according to your preferences.

For instance, you might use DataTag to differentiate Data based on their use, source, or purpose (like training or testing).

2. Creating, Editing, and Deleting DataTag

For creating, modifying, or removing a DataTag, you can access the Settings page of your Datalake and go to the Tags tab.

To create a new DataTag, simply click on Add new tag. This adds a new entry to the DataTag list, where you can provide a name and description for the new DataTag.

The DataTag is established when you click the + icon.

You can also change the name and description of an existing DataTag, or delete it using the trash icon after selecting it.

3. Attach DataTag to your Data

Once created, there are two ways to attach DataTag to your Data within the Datalake:

• Attaching DataTag when importing Data.
• Attaching DataTag to existing Data in your Datalake.

A. During Data Import into the Datalake

When you're importing Data into your Datalake (whether through direct upload or cloud storage scanning), you have the option to connect one or more DataTag to the incoming Data.

In this process, after selecting the files to upload, you can select the DataTag you want from the list.

You can also create new DataTag here, like this:

Once you click Create, the new DataTag gets added to the list and will be connected to the Data you're uploading when you click Upload.

B. Directly Within the Datalake

If you want to attach DataTag to Data that's already in your Datalake, simply select the images or videos you want to tag, and then choose Tag > Attach.

After that, you can link the selected Data to the DataTag you want from the list, and confirm by clicking Attach, please note that once again, remember that you can also create new DataTag from this attachment modal.

4. Visualizing DataTag on your Data

Now that you know how to create and attach DataTag to enhance your Datalake Organization, let's explore how to use these tags to effectively locate the specific Data you need within your Datalake.

You can begin by identifying the DataTag associated with each Data entry using the Table or Details views in your Datalake.

Searching for Data from their DataTag is the subject of the next section.

5. Search Data by DataTag

Once your DataTag are attached, three ways let you keep only the Data you are interested in.

A. With the Select Tags filter

Next to Select Date, above the list, the Select Tags button filters the Data on their DataTag without writing any query. Tick the DataTag you want and choose how they are combined, as detailed here:

• AND keeps the Data that carry all the selected DataTag, for instance the ones tagged both night and highway to build a targeted subset.
• OR keeps the Data that carry at least one of them, which is handy to gather the Data coming from several sources.
• NOT keeps the Data that carry none of the selected DataTag. Select the DataTag you use to flag unwanted Data (for instance duplicate or blurry) to display only the Data that none of them flags.

The number of Data displayed above the list follows the filter, and the filter is kept in the address of the page, so you can share the filtered list with a teammate by sharing the link.

For a Computer Vision team, the filtered list is the starting point of the curation: once it displays what you expect, select the Data and use the action buttons described here to attach or detach other DataTag, launch a Processing, or add them to a Dataset.

B. With the Query Language

A more SQL-like way to create filters, it allows to create more complex queries than with the filters only by combining them with filters on other properties related to the Data. It can be used together with the Select Tags filter, which then narrows down the result of your query.

C. From the Details view

Clicking one of the DataTag displayed in the Details view does the same as the filter in one gesture: the list is filtered on that DataTag, so you can go from the Data you are looking at to every other Data that shares its tag without opening the filter or writing a query.

6. Detach DataTag

If for any reason you need to detach one or several DataTag from a set of Data from your Datalake, you simply need to select the batch of Data that the DataTag needs to be detached from and click on Tag > Detach.

A modal will open listing all DataTag attached to any of the selected Data. You just need to click on the DataTag you want to detach from the Data selected and click on the Detach button.