In spite of remarkable success of the convolutional neural networks on semantic segmentation, they suffer from catastrophic forgetting: a significant performance drop for the already learned classes when new classes are added on the data, having no annotations for the old classes. We propose an incremental learning methodology, enabling to learn segmenting new classes without hindering dense labeling abilities for the previous classes, although the entire previous data are not accessible. The key points of the proposed approach are adapting the network to learn new as well as old classes on the new training data, and allowing it to remember the previously learned information for the old classes. For adaptation, we keep a frozen copy of the previously trained network, which is used as a memory for the updated network in absence of annotations for the former classes. The updated network minimizes a loss function, which balances the discrepancy between outputs for the previous classes from the memory and updated networks, and the mis-classification rate between outputs for the new classes from the updated network and the new ground-truth. For remembering, we either regularly feed samples from the stored, little fraction of the previous data or use the memory network, depending on whether the new data are collected from completely different geographic areas or from the same city.Our experimental results prove that it is possible to add new classes to the network, while maintaining its performance for the previous classes, despite the whole previous training data are not available.

Due to the various reasons such as atmospheric effects and differences in acquisition, it is often the case that there exists a large difference between spectral bands of satellite images collected from different geographic locations. The large shift between spectral distributions of training and test data causes the current state of the art supervised learning approaches to output unsatisfactory maps. We present a novel semantic segmentation framework that is robust to such shift. The key component of the proposed framework is Color Mapping Generative Adversarial Networks (ColorMapGAN), which can generate fake training images that are semantically exactly the same as training images, but whose spectral distribution is similar to the distribution of the test images. We then use the fake images and the ground-truth for the training images to fine-tune the already trained classifier. Contrary to the existing Generative Adversarial Networks (GANs), the generator in ColorMapGAN does not have any convolutional or pooling layers. It learns to transform the colors of the training data to the colors of the test data by performing only one element-wise matrix multiplication and one matrix addition operations. Thanks to the architecturally simple but powerful design of ColorMapGAN, the proposed framework outperforms the existing approaches with a large margin in terms of both accuracy and computational complexity.

The domain adaptation of satellite images has recently gained an increasing attention to overcome the limited generalization abilities of machine learning models when segmenting large-scale satellite images. Most of the existing approaches seek for adapting the model from one domain to another. However, such single-source and single-target setting prevents the methods from being scalable solutions, since nowadays multiple source and target domains having different data distributions are usually available. Besides, the continuous proliferation of satellite images necessitates the classifiers to adapt to continuously increasing data. We propose a novel approach, coined DAugNet, for unsupervised, multi-source, multi-target, and lifelong domain adaptation of satellite images. It consists of a classifier and a data augmentor. The data augmentor, which is a shallow network, is able to perform style transfer between multiple satellite images in an unsupervised manner, even when new data are added over the time. In each training iteration, it provides the classifier with diversified data, which makes the classifier robust to large data distribution difference between the domains. Our extensive experiments prove that DAugNet significantly better generalizes to new geographic locations than the existing approaches.