Xiaobai Training Camp Lesson 7 - Dataset
How to upload and preview dataset
In the previous lessons, we have learned to create a repository. After creating the repository, we can upload the corresponding dataset under the repository. When we are ready to run the repository on the cloud brain platform, the dataset comes from here.
So in this lesson, we will learn the main functions in the dataset, including uploading, previewing, setting and some precautions.
1. Go to the dataset page
First, after selecting the corresponding repository, open the [Datasets] page
2. Upload the dataset
Datasets can be uploaded by dragging files or clicking. The processing unit (CPU/GPU or Ascened NPU) should be selected before uploading
There are two buttons below the dataset name, [CPU/GPU] and [Ascend NPU], click these two buttons to switch between Cloud Brain 1 and Cloud Brain 2. Cloud Brain 1 provides CPU / GPU resources, Cloud Brain 2 provides Ascend NPU resources. You need to select the upload path according to specific needs.
The data set supports uploading in any format, but if you want to create a cloud brain task, the dataset format must be a zip compression format. In addition, cloud brain 1 and cloud brain 2's datasets are not shared, both computing platforms support the function of resuming upload from a breakpoint.
We can directly drag the file into it, or click to upload a data set in zip format, we can download some corresponding pictures on the Internet, and compress the folder with the compression software. When compressing, we need to select the zip format.
Complete file upload
In addition, we can also upload files in non-zip format, we choose a photo of a cat
After uploading, we can find that the zip format file not only has [copy download url] and [copy_md5] buttons, but also has [preview of the datasets] and [create label task] buttons.
3. Preview the dataset
For the dataset in zip format, click the file icon on the right to preview it
4. Dataset Settings
file is private by default, and you can click [Public] to make the file visible to everyone
For the convenience of viewing and managing datasets, there are labels on the right side of the page, such as dataset classification, research direction/application area, license, etc. Adding labels to datasets can improve the identification of items and improve retrieval rates.
There is an [edit] button on the right side of the dataset, you can edit the information of the dataset
After entering the editing page, you can add the dataset name and introduction, and click [Update Dataset] after entering the information.
Well, the corresponding functions of the data set are introduced here for you, and a general summary is:
- Upload the corresponding dataset according to the cloud brain task
- Preview datasets
- Permission and classification settings for datasets