An interpretable deep learning framework for land use and land cover (LULC) classification in remote sensing using Shapley additive explanations (SHAPs) is introduced. It utilizes a compact convolutional neural network (CNN) model for the classification of satellite images and then feeds the results to a SHAP deep explainer so as to strengthen the classification results. The proposed framework is applied to Sentinel-2 satellite images containing 27000 images of pixel size $64 \times 64$ and operates on three-band combinations, reducing the model’s input data by 77% considering that 13 channels are available, while at the same time investigating on how different spectrum bands affect predictions on the dataset’s classes. Experimental results on the EuroSAT dataset demonstrate the CNN’s accurate classification with an overall accuracy of 94.72 %, whereas the classification accuracy on three-band combinations on each of the dataset’s classes highlights its improvement when compared to standard approaches with larger number of trainable parameters. The SHAP explainable results of the proposed framework shield the network’s predictions by showing correlation values that are relevant to the predicted class, thereby improving the classifications occurring in urban and rural areas with different land uses in the same scene.

Over the past five years, interest in the literature regarding the security of the Internet of Medical Things (IoMT) has increased. Due to the enhanced interconnectedness of IoMT devices, their susceptibility to cyber-attacks has proportionally escalated. Motivated by the promising potential of AI-related technologies to improve certain cybersecurity measures, we present a comprehensive review of this emerging field. In this review, we attempt to bridge the corresponding literature gap regarding modern cybersecurity technologies that deploy AI techniques to improve their performance and compensate for security and privacy vulnerabilities. In this direction, we have systematically gathered and classified the extensive research on this topic. Our findings highlight the fact that the integration of machine learning (ML) and deep learning (DL) techniques improves both the performance of cybersecurity measures and their speed, reliability, and effectiveness. This may be proven to be useful for improving the security and privacy of IoMT devices. Furthermore, by considering the numerous advantages of AI technologies as opposed to their core cybersecurity counterparts, including blockchain, anomaly detection, homomorphic encryption, differential privacy, federated learning, and so on, we provide a structured overview of the current scientific trends. We conclude with considerations for future research, emphasizing the promising potential of AI-driven cybersecurity in the IoMT landscape, especially in patient data protection and in data-driven healthcare.

Detecting and localizing buildings is of primary importance in urban planning tasks. Automating the building extraction process, however, has become attractive given the dominance of Convolutional Neural Networks (CNNs) in image classification tasks. In this work, we explore the effectiveness of the CNN-based architecture U-Net and its variations, namely, the Residual U-Net, the Attention U-Net, and the Attention Residual U-Net, in automatic building extraction. We showcase their robustness in feature extraction and information processing using exclusively RGB images, as they are a low-cost alternative to multi-spectral and LiDAR ones, selected from the SpaceNet 1 dataset. The experimental results show that U-Net achieves a $91.9\%$ accuracy, whereas introducing residual blocks, attention gates, or a combination of both improves the accuracy of the vanilla U-Net to $93.6\%$, $94.0\%$, and $93.7\%$, respectively. Finally, the comparison between U-Net architectures and typical deep learning approaches from the literature highlights their increased performance in accurate building localization around corners and edges.

An efficient spectral band selection trustworthy Machine Learning (ML) framework for vineyard monitoring from UAV hyperspectral data is introduced. The UAV, equipped with Specim AFX-10, is used to capture data beyond the visible spectrum within the $400-1000$ nm wavelength range for a total of 224 bands. Popular supervised Machine Learning (ML) algorithms are utilized for detecting vegetation canopy in vineyards and distinguishing it from existing land uses, namely ground and shadow. Explainable AI (XAI) results accompany those from ML to identify the most important bands, and understand the contribution of their reflectance levels to the ML models. By doing so, the number of spectral bands is narrowed while maintaining the granularity of the HS data. Experimental results on UAVINE, a publicly available dataset within the context of this work, demonstrate excelling classification performance of Random Forest (RF) with an overall accuracy of 97.06%, and with precision, recall and F1-scores following accordingly. With the use of the computationally efficient Tree SHAP algorithm applied on the RF, the bands B106 (677nm- Red), B186 (897nm- NIR) and B211 (967nm- NIR) and B39 (498nm- Green) were identified as the most important ones to the model, enabling better visualization of the vineyard and band-based analysis for each one of the classes existing within the vineyard.

Non-scaling Stochastic Computing adder and subtracter architectures are introduced. They are modeled using Markov Chains to obtain important statistical properties enabling their design optimization. To demonstrate their efficacy, they are used to realize a stochastic computing-based image sharpening filter which is simulated in MATLAB and Synopsys. The filter’s computational efficiency is showcased with standard image processing metrics while its hardware resources are compared to those of the standard binary filter, highlighting the advantages of the proposed approach.

A stochastic computing (SC) adder architecture, based on sigma-delta modulation is introduced. The operation principle of the stochastic computing sigma-delta (SCSD) adder is presented and the adder is modeled using Markov Chains resulting in the derivation of output’s first-moment statistics and the demonstration of its fast-convergence. The SCSD’s single-bit output enables the connection to existing Stochastic Finite-State Machines realizing non-linear functions and the formation of cascade processing structures appropriate for efficient realizations of SC artificial neurons. To demonstrate the proposed adder’s efficacy, a SCSD adder-based neuron was designed and used as a building block of a SCSD Multi-Layer Perceptron (MLP). The computational accuracy of the SCSD MLP and the hardware resources it requires are compared to those of the Fixed-Point arithmetic implementation, highlighting the effectiveness and significant area savings of the proposed SC approach. Furthermore, the accuracy and hardware requirements of the proposed SCSD MLP are also compared to those of corresponding SC architectures in the literature.