Folders
We explain how to make data easily searchable, by organizing your ML datasets with folders and tags.
Without Folders in place - particularly for large, sprawling enterprises - your data will become increasingly difficult to manage. In this Darwin Fundamentals session, we tackle Folders, and outline how you can use them to simplify and categorize your datasets.
For teams frequently annotating thousand of datasets, a clean and organized approach to data is paramount to success. That’s why Folders are a particularly useful feature within V7’s Darwin, allowing you to easily organize, manage, and navigate through your data.
We tackle how to create new folders, how to add data to these folders, and how to “nest” folders inside other folders to create a directory structure.
We also showcase how to toggle Folders on and off as a view, to allow Users to easily switch between seeing all images and videos together or filtering based on folder organization.
Finally, we outline what Folders can, and can’t do - including the fact that your Folders will have no impact on Exports and will not restrict annotators from labeling images within them.
You’ll come away from this video with a detailed understanding of how to create a well-structured dataset to enhance your annotation and classification workflows. With this feature, you’ll understand how to seamlessly manage data, making image and annotation tasks more streamlined in the process.
Press the folder icon on the top right to activate a direct review. You won't just be seeing images and video. Now, you'll also see folders. Click on an image or video or shifts. Click to select more than one and press the add to folder button at the bottom to create a new folder. Simply type whatever you want to name it and press this folder icon or enter the images you had selected will now be there. You can also nest folders inside of other folders. To create a directory structure, just select images from anywhere on the dataset. It doesn't have to be from a parent folder. You can help yourself by making the icons bigger and selecting the images you need. Then again, add to folder.
This time we will type in a path burger slash fries. This will create a fries folder inside the burger folder. If the burger folder didn't exist beforehand, it would've created one. You can view the nested structure of your folder by clicking on the add to folder button. Clicking on any of these entries will add your images to them.
You can also send images back to the root directory by clicking on the top. Entry. Folders can be toggled on and off as a view, so if you prefer see all images and video at the same time and filter from there. You can always turn this off. It's a handy feature to replicate the directory structure of legacy data sets or to organize data with labels that wouldn't serve classification purposes, such as camera views or batches.
And you can always use them to bundle data and classify it from there. Note that folders don't affect your exports in any way, nor do they prevent annotators from labeling images within them.
