Unsupervised {Spectral}-{Spatial} {Feature} {Learning} via {Deep} {Residual} {Conv}-{Deconv} {Network} for {Hyperspectral} {Image} {Classification}

Supervised approaches classify input data using a set of representative samples for each class, known as training samples. The collection of such samples is expensive and time demanding. Hence, unsupervised feature learning, which has a quick access to arbitrary amounts of unlabeled data, is conceptually of high interest. In this paper, we propose a novel network architecture, fully Conv-Deconv network, for unsupervised spectral-spatial feature learning of hyperspectral images, which is able to be trained in an end-to-end manner. Specifically, our network is based on the so-called encoder-decoder paradigm, i.e., the input 3-D hyperspectral patch is first transformed into a typically lower dimensional space via a convolutional subnetwork (encoder), and then expanded to reproduce the initial data by a deconvolutional subnetwork (decoder). However, during the experiment, we found that such a network is not easy to be optimized. To address this problem, we refine the proposed network architecture by incorporating: 1) residual learning and 2) a new unpooling operation that can use memorized max-pooling indexes. Moreover, to understand the “black box,” we make an in-depth study of the learned feature maps in the experimental analysis. A very interesting discovery is that some specific “neurons” in the first residual block of the proposed network own good description power for semantic visual patterns in the object level, which provide an opportunity to achieve “free” object detection. This paper, for the first time in the remote sensing community, proposes an end-to-end fully Conv-Deconv network for unsupervised spectral-spatial feature learning. Moreover, this paper also introduces an in-depth investigation of learned features. Experimental results on two widely used hyperspectral data, Indian Pines and Pavia University, demonstrate competitive performance o- tained by the proposed methodology compared with other studied approaches.     «

Supervised approaches classify input data using a set of representative samples for each class, known as training samples. The collection of such samples is expensive and time demanding. Hence, unsupervised feature learning, which has a quick access to arbitrary amounts of unlabeled data, is conceptually of high interest. In this paper, we propose a novel network architecture, fully Conv-Deconv network, for unsupervised spectral-spatial feature learning of hyperspectral images, which is able to...     »

object detection, image classification, Support vector machines, Training, learning (artificial intelligence), hyperspectral image classification, Hyperspectral imaging, Convolutional network, deconvolutional network, deep residual Conv-Deconv network, end-to-end manner, Feature extraction, input 3-D hyperspectral patch, learned feature maps, learned features, network architecture, Network architecture, residual learning, unsupervised feature learning, unsupervised learning, unsupervised spectral-spatial feature learning, unsupervised spectral–spatial feature learning, widely used hyperspectral data     «

object detection, image classification, Support vector machines, Training, learning (artificial intelligence), hyperspectral image classification, Hyperspectral imaging, Convolutional network, deconvolutional network, deep residual Conv-Deconv network, end-to-end manner, Feature extraction, input 3-D hyperspectral patch, learned feature maps, learned features, network architecture, Network architecture, residual learning, unsupervised feature learning, unsupervised learning, unsupervised spectra...     »