Search for: All records

Implementing local contextual guidance principles in a single-layer CNN architecture, we propose an efficient algorithm for developing broad-purpose representations (i.e., representations transferable to new tasks without additional training) in shallow CNNs trained on limited-size datasets. A contextually guided CNN (CG-CNN) is trained on groups of neighboring image patches picked at random image locations in the dataset. Such neighboring patches are likely to have a common context and therefore are treated for the purposes of training as belonging to the same class. Across multiple iterations of such training on different context-sharing groups of image patches, CNN features that are optimized in one iteration are then transferred to the next iteration for further optimization, etc. In this process, CNN features acquire higher pluripotency, or inferential utility for any arbitrary classification task. In our applications to natural images and hyperspectral images, we find that CG-CNN can learn transferable features similar to those learned by the first layers of the well-known deep networks and produce favorable classification accuracies. 

The segmentation of sky images into regions of cloud and clear sky allows atmospheric scientists to determine the fraction of cloud cover and the distribution of cloud without resorting to subjective estimates by a human observer. This is a challenging problem because cloud boundaries and cirroform cloud regions are often semi-transparent and indistinct. In this study, we propose a lightweight, unsupervised methodology to identify cloud regions in ground-based hemispherical sky images. Our method offers a fast and adaptive approach without the necessity of fixed thresholds by utilizing K-means clustering on transformed pixel values. We present the results of our method for two data sets and compare them with three different methods in the literature.