Deep learning (DL) techniques are becoming important to solve a number many of image processing tasks. Among common algorithms, the convolutional neural network and recurrent neural networkbased systems achieves the stateofthe art results on satellite and aerial imagery in many applications. While these approaches are subjected to the scientific interest, there is currently a no operational and generic implementation available at the user level for the remote sensing (RS) community. In this paper, we propose a framework whichenablesthe use of DL techniques with RS images and geospatial data. The results takes roots in two extensively used opensource libraries namely, the RS image processing library Orfeo ToolBox and the highperformance numerical computation library TensorFlow. Though ,it can be capable to apply deep nets without restriction on image size and is found computationally efficient, regardless of hardware configuration.
展开▼