Core Sampling Framework for Pixel Classification
This work addresses pixel classification for image segmentation, particularly in SAR imagery, but appears incremental as it builds on existing transfer learning and feature combination techniques.
The authors tackled pixel-level image understanding by developing a core sampling framework that combines activation maps from pretrained CNNs with test data features, processed through a Deep Belief Network. They demonstrated segmentation improvements on SAR and CAMVID datasets using a VGG-16 model, though no specific numerical results were provided.
The intermediate map responses of a Convolutional Neural Network (CNN) contain information about an image that can be used to extract contextual knowledge about it. In this paper, we present a core sampling framework that is able to use these activation maps from several layers as features to another neural network using transfer learning to provide an understanding of an input image. Our framework creates a representation that combines features from the test data and the contextual knowledge gained from the responses of a pretrained network, processes it and feeds it to a separate Deep Belief Network. We use this representation to extract more information from an image at the pixel level, hence gaining understanding of the whole image. We experimentally demonstrate the usefulness of our framework using a pretrained VGG-16 model to perform segmentation on the BAERI dataset of Synthetic Aperture Radar(SAR) imagery and the CAMVID dataset.