Prompt Segmentation
Instructions for use
Prompt-based segmentation using the SAM model involves using vector data as prompt information to segment images. SuperMap currently offers no-prompt, rectangular-box prompt, and polygon prompt segmentation capabilities, enabling precise segmentation of targets within images.
Parameter Description
| Parameter name | Default | Parameter description | Parameter Type |
|---|---|---|---|
| Dataset | Image Target Dataset in Select Works pace. | DatasetImage | |
| Prompt type | No prompt | Three types are supported: "Rectangular box prompt, Polygon prompt, and no Prompt data". The default is no prompt. | String |
| Prompt data | Support Vector Region Dataset. | DatasetVector | |
| Model magnitude | 4 | The larger the number of parameters is, the more complex the model is, and the higher the accuracy of Split Result is. | String |
| Sampling mode | Equidistant sampling and Local similarity sampling are supported. When prompted that the Data Type is "No prompt", Equidistant sampling is the default; When Prompt Data is "Polygon prompt" ", the default Local similarity sampling; when Prompt Data Type is" Rectangular box prompt "", this parameter is not optional. | Enum | |
| The size of tiles | 1024 | The size of the picture entered into the model during inference. The default is 1024. | Integer |
| Tile overlap (pixels): | 0 | The default value is 0 to reduce the problem of insufficient prediction of Tile edge data. The larger the Tile overlap, the longer it takes to infer the entire image. | Integer |
| Amount of single step operation | 1 | The number of pictures processed at one time during inference. The default value is 1. Within a reasonable range, the amount of single-step computation is directly proportional to the memory (video memory) occupied, and inversely proportional to the reasoning time. | Integer |
| Processor type | A central proces unit (CPU) or a graphics proces unit (GPU) of a computer may be used to process data. And that GPU has high calculation speed. | String | |
| GPU number | 0 | Specifies the GPU identification number that processes the data. The default value is 0. If multiple cards are used, separate them with "?". When the processor is a CPU, this parameter is -1 and is not optional. | String |
| Result Datasource | The Datasource where the specified Dataset storing the result is located. | DataSource | |
| Result Dataset | resultDataName | The output Result Dataset is Vector Data. | String |
Output Result
| Parameter name | Parameter description | Parameter Type |
|---|---|---|
| Prompt segmentation result | Output Vector Dataset | DatasetVector |