Incremental Learning for Semantic Segmentation of Large-Scale Remote Sensing Data

Paper

The paper can be found here.


Additional Results

For the sake of more compact visualization in visual results of the paper, instead of multiple binary predicted maps, multi-class predictions are provided by assigning each pixel to the class, for which the highest probability is produced (see Fig. 9).

The original binary maps by only our approach for the close-ups from the Luxcarta dataset are illustrated in the figure below:


Source Code

If you are interested in the code, please fill in the form below:

Comments are closed.