Feature Description
Super-resolution reconstruction refers to converting low-resolution images into high-resolution images through super-resolution technology. The core principle of deep learning-based super-resolution reconstruction is to train deep neural networks to learn the mapping relationship between low-resolution and high-resolution images, thereby predicting and supplementing missing details to enhance image clarity, sharpness, and texture representation.
The input source data consists of low-resolution images, and the output results are corresponding high-resolution images. The super-resolution scaling factor is determined by the input model file. For example, a 10,000×10,000 pixel image will output a 20,000×20,000 pixel image when using a 2× super-resolution model, and a 40,000×40,000 pixel image when using a 4× super-resolution model.
Parameter Description
| Parameter | Description | Type |
|---|---|---|
| File Type | Select whether the file type is a dataset or folder. | String |
| Datasource | File-based datasource containing the image dataset. | String |
| Dataset | Select an image dataset or mosaic dataset. | DatasetVector |
| File Path | Select a folder path to automatically read image files. Supports five image types: TIF/TIFF/IMG/PNG/JPG (case-insensitive). | DatasetVector |
| Model File | Specify the model file (*.sdm) for super-resolution reconstruction. | String |
| Batch Size | Number of images processed per training step. Increasing this value improves efficiency but is limited by GPU memory (GPU inference) or RAM (CPU inference). | Integer |
| Processor Type | Specify the processor type. | String |
| GPU ID | Specify the GPU identifier for data processing. Default: GPU. | String |
| Result Save Path | Folder path to save results. | String |
| Result Filename | Specify the output filename. For folder-type source data, this name will be appended with "_" to original filenames to distinguish reconstruction results. | String |
Output
| Parameter | Description | Type |
|---|---|---|
| Super-Resolution Reconstruction Result | Output format is TIFF for dataset-type sources, and matches original format for folder-type sources. | String |