No abstract

The systematic monitoring of the Earth using optical satellites is limited by the presence of clouds. Accurately detecting these clouds is necessary to exploit satellite image archives in remote sensing applications. Despite many developments, cloud detection remains an unsolved problem with room for improvement, especially over bright surfaces and thin clouds. Recently, advances in cloud masking using deep learning have shown significant boosts in cloud detection accuracy. However, these works are validated in heterogeneous manners, and the comparison with operational threshold-based schemes is not consistent among many of them. In this work, we systematically compare deep learning models trained on Landsat-8 images on different Landsat-8 and Sentinel-2 publicly available datasets. Overall, we show that deep learning models exhibit a high detection accuracy when trained and tested on independent images from the same Landsat-8 dataset (intra-dataset validation), outperforming operational algorithms. However, the performance of deep learning models is similar to operational threshold-based ones when they are tested on different datasets of Landsat-8 images (inter-dataset validation) or datasets from a different sensor with similar radiometric characteristics such as Sentinel-2 (cross-sensor validation). The results suggest that (i) the development of cloud detection methods for new satellites can be based on deep learning models trained on data from similar sensors and (ii) there is a strong dependence of deep learning models on the dataset used for training and testing, which highlights the necessity of standardized datasets and procedures for benchmarking cloud detection models in the future.

No abstract

The number of Earth observation satellites carrying optical sensors with similar characteristics is constantly growing. Despite their similarities and the potential synergies among them, derived satellite products are often developed for each sensor independently. Differences in retrieved radiances lead to significant drops in accuracy, which hampers knowledge and information sharing across sensors. This is particularly harmful for machine learning algorithms, since gathering new ground truth data to train models for each sensor is costly and requires experienced manpower. In this work, we propose a domain adaptation transformation to reduce the statistical differences between images of two satellite sensors in order to boost the performance of transfer learning models. The proposed methodology is based on the Cycle Consistent Generative Adversarial Domain Adaptation (CyCADA) framework that trains the transformation model in an unpaired manner. In particular, Landsat-8 and Proba-V satellites, which present different but compatible spatio-spectral characteristics, are used to illustrate the method. The obtained transformation significantly reduces differences between the image datasets while preserving the spatial and spectral information of adapted images, which is hence useful for any general purpose cross-sensor application. In addition, the training of the proposed adversarial domain adaptation model can be modified to improve the performance in a specific remote sensing application, such as cloud detection, by including a dedicated term in the cost function. Results show that, when the proposed transformation is applied, cloud detection models trained in Landsat-8 data increase cloud detection accuracy in Proba-V.

Automatic and reliable classification of images of white blood cells is desirable for inexpensive, quick and accurate health diagnosis worldwide. In contrast to previous approaches which tend to rely on image segmentation and a careful choice of ad hoc (geometric) features, we explore the possibilities of local image descriptors, since they are a simple approachthey require no explicit segmentation, and yet they have been shown to be quite robust against background distraction in a number of visual tasks. Despite its potential, this methodology remains unexplored for this problem. In this work, images are therefore characterized with the well-known visual bag-of-words approach. Three keypoint detectors and five regular sampling strategies are studied and compared. The results indicate that the approach is encouraging, and that both the sparse keypoint detectors and the dense regular sampling strategies can perform reasonably well (mean accuracies of about 80% are obtained), and are competitive to segmentation-based approaches. Two of the main findings are as follows. First, for sparse points, the detector which localizes keypoints on the cell contour (oFAST) performs somehow better than the other two (SIFT and CenSurE). Second, interestingly, and partly contrary to our expectations, the regular sampling strategies including hierarchical spatial information, multi-resolution encoding, or foveal-like sampling, clearly outperform the two simpler uniform-sampling strategies considered. From the broader perspective of expert and intelligent systems, the relevance of the proposed approach is that, since it is very general and problem-agnostic, it makes unnecesary human expertise to be elicited in the form of explicit visual cues; only the labels of the cell type are required from human domain experts.