Deep learning has achieved unprecedented accuracy in a variety of fields, including remote sensing. But not every remote sensing professional has a strong programming background, which is the required skill for practising deep learning. While enough material and sample code for deep learning applications are available in online repositories, pre-processing the remote sensing datasets to reach the training stage is still a tedious task and requires programming knowledge. Proprietary options to pre-process the data only on mouse clicks are available. However, access to such licensed software is often expensive and hence limited for a large population of researchers and practitioners. The plugin bridges the gap by enabling users to generate training samples in the form of image chips for deep learning applications.

Curious to understand functionalities of the plugin? read here. The plugin can be downloaded from inside the QGIS application, make sure you have checked the experimental plugin box in the settings. Feel free to report bugs and get support. Suggestions for additional features are most welcomed.
Email: [email protected]

The below screenshot shows generated training samples for identifying industrial buildings in high-resolution (1.6m) satellite image.

NOTE:

• The plugin was originally developed on a Windows 8.1 operating system.
• If you screen resolution is too high and you can not see the plugin window properly, you may want to change the QGIS scaling factor in your computer's environment path variables. For windows you can use this from the command prompt shell:

setx QT_AUTO_SCREEN_SCALE_FACTOR 2

Later if you wish to roll back, set the scale factor back to one using this:

setx QT_AUTO_SCREEN_SCALE_FACTOR 1