Skip to main content

Creating a new dataset

Dataset creation flow#

To create a new dataset, click the + New dataset button in the Datasets view. This initiates the dataset creation flow, a simple step-by-step process for turning your data files into datasets.

1. Enter details#

In the 'Enter details' screen, enter a dataset title and optional description. A clear title and description helps keep your data organized and helps collaborators find what they are looking for. When you are happy, click Next.

2. Upload data#

The 'Upload data' screen allows you to upload your data files.

You can upload files directly from your computer or integrate with your own private cloud data provider by toggling the Private cloud toggle.

Upload videos from a computer#

Click the upload rectangle or drag video files into it to add your files. Once uploaded, files will appear in the Dataset data section on the right. Successful uploads will have a green tick; unsuccessful one will have a red cross.

Upload images from a computer#

Click on the Image upload button to launch the 'New image group' pop-up.

Images in Encord are organised into 'Image groups'. An image group is a collection images compressed into a sequence of frames, allowing for more efficient annotation of semantically similar images. You can still label each image individually, however.

Click the upload rectangle or drag image files into it to add your files. Once you are happy with your image group, click the Create image group button. Similar to video files, files will appear in the Dataset data section on the right once uploaded.

note

Images in a group are all assumed to be the same resolution. If the resolutions of the images don't match, they will be separated into different groups of matched resolutions.

Connect to private cloud storage#

You can also link data that you have stored in a cloud provider by toggling the Private cloud toggle above the upload section.

We provide integrations for Azure, Google Cloud Platform, Amazon's S3 and Open Telekom Cloud. Please check the Private Cloud Integration section on how to integrate your cloud data.

3. Create the dataset#

Once you have uploaded or connected all your data, click the Create dataset button to create your dataset. This will now appear as a row in the Datasets view with the appropriate title and description.

Supported data formats#

We support all the major video and image formats. However not all video formats are supported by all browsers for playback. If you have trouble with playback in the Label editor, please refer to the documentation for your browser.

For video data, the Label editor expects integer frame rates from 10-30 fps. Non-integer frame rates will generate a warning in the Label editor. Although labeling can continue with these frame rates, please be aware this can introduce errors into your annotations.