Step 5, Datasets – Importing Data into dataspan.ai

Importing a Dataset

To import a Dataset:

1. In the Project page, click the Datasets option.

2. Click the IMPORT DATASET button. The following displays:

3. Step 1: Storage

Specify the name of the Dataset and the storage from which it will be imported.

Note: Currently, only AWS S3 is allowed. Support for other storage options is coming soon.

Datasets must each be in a separate sub-folder in your storage (S3) in order to be imported into separate Datasets in dataspan.ai.

Tip: It’s not necessary, but you may find it helpful to import the images of the foreground objects (the What to insert, such as defects) as a separate Dataset from the images of the background canvases (the Where to insert).

Click the NEXT STEP button. The following displays:

4. Step 2: Dataset Type

Specify the type of Dataset and metadata (annotations), as follows:

Image Folder

This is the simplest kind of Dataset. It is a folder of images without any metadata, such as bounding boxes and segmentation masks. The following displays:

Note: There are no additional fields to fill in. Skip to Step ‎5 in this procedure to click the click the IMPORT DATASET button. ++++

Bounding Boxes

Specifies that this Dataset includes an annotation file(s) defining the bounding boxes for some or all of the images: Pascal VOC (which uses an XML annotation file per image), COCO (which uses a single JSON annotation file) or YOLO (which also uses a TXT annotation file per image). The following displays:

Segmentation Masks

Specifies that this Dataset includes metadata of segmentation masks for some or all of the images. Grayscale or multi-class color-coded mask files can act as segmentation masks. The following displays:

Tip: If some images have associated bounding box files and other images have associated segmentation masks, then import two Datasets, and in the Choose Dataset type field select Bounding Box for one and Segmentation Mask for the other.

1. Click the IMPORT DATASET button. While the Datasets is being imported and analyzed, the Datasets page is displayed. The new Dataset will be added there and become clickable after it is ready. For example, in the picture below, the Dataset on the right is still being processed, and the two Datasets on the left are ready:

1. We now recommend reviewing the images in the imported Dataset, as described in Reviewing Datasets.