Supervisely

Use case: upload project in Supervisely format

For this case the structure of directory should be the following:
1
my_project
2
โ”œโ”€โ”€ meta.json
3
โ”œโ”€โ”€ dataset_name_01
4
โ”‚ โ”œโ”€โ”€ ann
5
โ”‚ โ”‚ โ”œโ”€โ”€ img_x.jpeg.json
6
โ”‚ โ”‚ โ”œโ”€โ”€ img_y.jpeg.json
7
โ”‚ โ”‚ โ””โ”€โ”€ img_z.jpeg.json
8
โ”‚ โ””โ”€โ”€ img
9
โ”‚ โ”œโ”€โ”€ img_x.jpeg
10
โ”‚ โ”œโ”€โ”€ img_y.jpeg
11
โ”‚ โ””โ”€โ”€ img_z.jpeg
12
โ”œโ”€โ”€ dataset_name_02
13
โ”‚ โ”œโ”€โ”€ ann
14
โ”‚ โ”‚ โ”œโ”€โ”€ img_x.jpeg.json
15
โ”‚ โ”‚ โ”œโ”€โ”€ img_y.jpeg.json
16
โ”‚ โ”‚ โ””โ”€โ”€ img_z.jpeg.json
17
โ”‚ โ””โ”€โ”€ img
18
โ”‚ โ”œโ”€โ”€ img_x.jpeg
19
โ”‚ โ”œโ”€โ”€ img_y.jpeg
20
โ”‚ โ””โ”€โ”€ img_z.jpeg
Copied!
Directory "my_project" contains two folders and file meta.json. For each folder a corresponding dataset will be created inside the project. As you can see, images are separated from annotations.
Let's consider the example above. To upload that project user has to select meta.json file and two directories (dataset_name_01 and dataset_name_02) and drag-and-drop these three elements.
We recomemnd to split images logically into several datasets inside one project.

Read more about Supervisely format

The full specification for the format can be found here.
Last modified 1yr ago