Export Sample Library

Model Training needs to be trained based on the sample library. After the samples are drawn, they can be exported as the sample library.

Once Sample Management is ready, it can be exported as a sample library that meets the sample data requirements of Model Training to provide sample data for Model Training. The main parameters of Export Sample Library are as follows:

Sample Library Name: Set the Sample Library Name. The default is the Sample Library Name at the time of creation.
File Path: Set the Save path of the sample library.
Sample Library Purpose: If it is the sample library of Object Detection, three sample library purposes are provided: Object Detection, Semantic Segmentation, and Sample Segmentation.
- Object Detection: The sample identifies the object and location in the image, usually using Bounds, as shown in Figure 2 below;
- Semantic Segmentation: The polygon is usually used to identify the edge of the sample. It is necessary to distinguish each pixel in the figure, as shown in Figure 3 below. It is enough to mark it as human, sheep, dog and grass.
- Instance Segmentation: Use polygons to precisely identify the edge of the object, and mark different individuals of the same object in the image, as shown in Figure 4 below, which need to be marked as Sheep 1, Sheep 2, Sheep 3, Sheep 4 and Sheep 5.
Image Classification Sample Library: In the Image Classification Sample Library, according to the type of the sample image, the image is stored in the folder named by the category index, and the CSV, txt, SDA and other files of the Samples Category information are output. The following figure shows the sample library results of the ammeter image Samples Category.
Object Detection Sample Library: The Object Detection Sample Library records the sample image, the type and size of the label, and the Config File of the sample library.

Related topics