doc-exports/docs/modelarts/umn/dataprepare-modelarts-0011.html

Introduction to Importing Data from OBS

Import Modes

You can import data from OBS through an OBS path or a manifest file.

• OBS path: indicates that the dataset to be imported has been stored in an OBS path. In this case, select an OBS path that you can access. In addition, the directory structure in the OBS path must comply with the specifications. For details, see Specifications for Importing Data from an OBS Directory. This import mode is available only for the following types of datasets: Image classification, Object detection, Text classification, Table, and Sound classification. For other types of datasets, data can be imported only through a manifest file.
• Manifest file: indicates that the dataset file is in the manifest format and the manifest file has been uploaded to OBS. The manifest file defines the mapping between labeling objects and content. For details about the specifications of manifest files, see Specifications for Importing a Manifest File.

Before importing an object detection dataset, ensure that the labeling range of the labeling file does not exceed the size of the original image. Otherwise, the import may fail.

Table 1 Import modes supported by datasets

Dataset Type	Labeling Type	From an OBS Path	From a Manifest File
Image	Image classification	Supported You can import unlabeled or labeled data. Format specifications of labeled data: Image classification	Supported You can import unlabeled or labeled data. Format specifications of labeled data: Image classification
	Object detection	Supported You can import unlabeled or labeled data. Format specifications of labeled data: Image classification	Supported You can import unlabeled or labeled data. Format specifications of labeled data: Object detection
	Image segmentation	Supported You can import unlabeled or labeled data. Format specifications of labeled data: Object detection	Supported You can import unlabeled or labeled data. Format specifications of labeled data: Object detection
Audio	Sound classification	Supported You can import unlabeled or labeled data. Follow the format specifications described in Sound classification.	Supported You can import unlabeled or labeled data. Format specifications of labeled data: Sound classification
	Speech labeling	Supported You can import unlabeled data.	Supported You can import unlabeled or labeled data. Format specifications of labeled data: Speech labeling
	Speech paragraph labeling	Supported You can import unlabeled data.	Supported You can import unlabeled or labeled data. Format specifications of labeled data: Speech paragraph labeling
Text	Text classification	Supported You can import unlabeled or labeled data. Format specifications of labeled data: Text classification	Supported You can import unlabeled or labeled data. Format specifications of labeled data: Text classification
	Named entity recognition	Supported You can import unlabeled data.	Supported You can import unlabeled or labeled data. Format specifications of labeled data: Named Entity Recognition
	Text triplet	Supported You can import unlabeled data.	Supported You can import unlabeled or labeled data. Format specifications of labeled data: Text triplet
Video	Video	Supported You can import unlabeled data.	Supported You can import unlabeled or labeled data. Format specifications of labeled data: Video Labeling
Other	Free format	Supported You can import unlabeled data.	-
Tables	Tables	Supported Follow the format specifications described in Tables.	-

Parent topic: Importing Data from OBS