Plant diseases have been detrimental for the agriculture industry, as they cause substantial crop loss globally. To overcome this, IoT and AI-based smart agriculture solutions are being deployed for plant disease detection. However, a diverse range of crops and their diseases pose enormous challenges to these methods. Additionally, limited generalizability and the black-box nature of existing deep learning models, together with the scarcity of in-field datasets, are the main bottlenecks in developing efficient and acceptable solutions for large-scale applications. In the present work, a lightweight model 'ConViTX' is proposed for plant disease classification that demonstrates improved generalizability and explainability. The compact architecture of ConViTX uses a fusion of convolutional neural networks and vision transformers to simultaneously capture local and global features. Remarkably, ConViTX outperforms nine state-of-the-art deep learning methods on four publicly available datasets and a self-collected in-field maize dataset. Furthermore, the model demonstrates explainable prediction through Gradient Weighted Class Activation Maps and Locally Interpretable Model-Agnostic Evaluations. ConViTX attains 98.8% accuracy on the maize dataset and 61.42% on drone camera-captured raw images. With only 0.7 million parameters and 0.647 billion operations per second, the proposed model has the potential for deployment on resource-constrained precision agriculture setups.

The increasing frequency and intensity of wildfires in recent years have not only devastated forest ecosystems, but have also caused a significant economic burden. According to a World Economic Forum report, annual expenditures to combat wildfire hazards is estimated to be more than$ {\$} 50$billion. This calls for advanced solutions, such as remote sensing surveillance and the use of artificial intelligence for wildfire management. In recent years, several vision-based artificial intelligence techniques have been proposed for fire–smoke image classification that utilise convolutional neural networks. However, challenges persist, particularly in identifying fire–smoke under complex atmospheric conditions. In this article, we introduce a novel multiattention network that interlaces the vision transformer and convolutional neural network to detect fire–smoke in diverse conditions, including clouds, fog, hurricanes, storms, snow, and normal weather. The proposed model not only outperforms eight state-of-the-art fire–smoke image classification methods, but also reduces false alarms by 30% on IIITDMJ$\_$Smoke dataset and by 6% on UTSC$\_$SmokeRS dataset. The model also efficiently identifies even tiny occurrence of smoke covering as little as 2% area of an image. The model has also been tested on industrial chimney smoke images and outdoor video fire–smoke scenes. Furthermore, the lightweight architecture of the model with only 0.7 million parameters and 0.2 billion floating point operations per second makes it suitable for deployment on Internet of Things-based forest and industrial surveillance systems.

Extraction of retinal vascular parts is an important task in retinal disease diagnosis. Precise segmentation of the retinal vascular pattern is challenging due to its complex structure, overlapping with other anatomical structures, and crucial thin vascular structures. In recent years, complex and heavy deep learning networks have been proposed to segment retinal blood vessels accurately. However, these methods fail to detect the thin vascular structure among different patterns of thick vessels. An attention-based novel architecture is proposed to segment the thin vasculature to address this limitation. The proposed model comprises a shallow U-Net based encoder-decoder architecture with split-fuse attention (SFA) block. The proposed SFA block enables the network to identify the placement of pixels for the tree-shaped vessel patterns at their relative position during the reconstruction phase in the decoder. The attention block aggregates low-level and high-level semantic information, improving the vessel segmentation performance. Experimentation performed on publicly available fundus datasets, DRIVE, HRF, CHASE-DB1, and STARE show that the proposed method performs better than the current state-of-the-art methods. The results demonstrate the adaptability of the proposed model for clinical applications due to its low memory footprint and better performance.

This work proposes a novel application of digital twins in the field of biometric security. Biometric systems have become widespread but their use carries risks of privacy invasion attacks due to the sensitive nature of biometric data. To address these concerns, we propose creating biometric clones for digital access and authentication systems. A user's fingerprint can act as a virtual representation or cyberproxy, allowing users to exist in the digital world with a unique, changeable, and privacy-preserving identity. The digital twin or clone fingerprint is generated using deep neural networks combined with a user-specific token/key. This approach allows third parties to process and store the proxy biometrics without putting the user's personal information at risk. We suggest that this approach could provide a safer and more secure alternative to traditional biometric security systems.

Alzheimer’s disease is an incurable neurodegenerative disease primarily affecting the elderly population. Efficient automated techniques are needed for early diagnosis of Alzheimer’s. Many novel approaches are proposed by researchers for classification of Alzheimer’s disease. However, to develop more efficient learning techniques, better understanding of the work done on Alzheimer’s is needed. Here, we provide a review on 165 papers from 2005 to 2019, using various feature extraction and machine learning techniques. The machine learning techniques are surveyed under three main categories: support vector machine (SVM), artificial neural network (ANN), and deep learning (DL) and ensemble methods. We present a detailed review on these three approaches for Alzheimer’s with possible future directions.

The cancelable biometric-based template protection method proposed in this paper addresses security and privacy concerns emerging from the phenomenal usage of biometric systems. Cancelable biometric transforms the original biometric identity of a user to a pseudo-biometric identity that is used for storage and matching purposes. The use of pseudo-identity mitigates privacy risks and allows revocability in case of compromise. This paper proposes a novel template transformation technique named random distance method which not only generates discriminative and privacy preserving revocable pseudo-biometric identities, but also reduces their size by 50%. Extensive experimentation is performed to analyze recognition and protection performance on unimodal and multimodal pseudo-identities generated with various biometric modalities such as face, thermal face, palmprint, palmvein, and fingervein. It is observed that the matching performance obtained with the proposed cancelable templates in the worst-case is closer to the performance achieved in the original domain. Also, multimodal cancelable biometric templates generated with the proposed method are observed for improved performance. Furthermore, the proposed approach is successfully analyzed for non-invertibilty, unlinkability, as well as its resistance for various types of attacks like attacks via record multiplicity, dictionary, false accepts, and brute force.

Mammogram enhancement is a key step to detect breast cancer using digital mammogram. The present study investigates mesh‐free radial basis function (RBF) collocation method to solve linear diffusion equation for image enhancement of mammograms. The proposed algorithm is compared with the mesh‐based finite difference method as well as other existing enhancement methods such as unsharp masking, histogram equalisation, and contrast limited adaptive histogram equalisation. Specifically, figure‐of‐merits with emphasis on image contrast and computational time are assessed and compared with different image processing techniques. The proposed algorithm has been applied towards the enhancement of all 322 sample mammogram images of Mini‐Mammographic Image Analysis Society and randomly selected 300 sample images from Digital Database for Screening Mammography databases. Finally, mean and standard deviation of contrast improvement index and peak signal‐to‐noise ratio have been estimated and presented. The outcome of this study indicates that the mesh‐free‐based RBF collocation method proves to be a computationally efficient and valuable contrast enhancement technique in an area of mammography.

An illumination and expression invariant face recognition method based on uniform local binary patterns (uLBP) and Legendre moments is proposed in this work. The proposed method exploits uLBP texture features and Legendre moments to make a feature representation with enhanced discriminating power. The input images are preprocessed to extract the face region and normalized. From normalized image, uLBP codes are extracted to obtain texture image which overcomes the effect of monotonic temperature changes. Legendre moments are computed from this texture image to get the required feature vector. Legendre moments conserve the spatial structure information of the texture image. The resultant feature vector is classified using k-nearest neighbor classifier with [Formula: see text] norm. To evaluate the proposed method, experiments are performed on IRIS and NVIE databases. The proposed method is tested on both visible and infrared images under different illumination and expression variations and performance is compared with recently published methods in terms of recognition rate, recall, length of feature vector, and computational time. The proposed method gives better recognition rates and outperforms other recent face recognition methods.

Hand-gesture-based commands may replace touch and electromechanical input panels to make public interactive response systems (IRS) more accessible. This work presents a prototype framework for vision-based mid-air unistroke character input, which can be adapted as an interface for the IRS. At first, we developed an acquisition module that effectively spots the legitimate gesture trajectory by implementing pen-up and pen-down actions using depth thresholding and velocity tracking. The extracted trajectory is recognized through a novel, fast, and easy to implement the equipolar signature (EPS) technique. Apart from resistance to rotation, scale, and translation variations, EPS exhibits neutrality to stroking directions as well. On the three selfcollected datasets comprising of digits, alphabets, and symbols, the EPS scheme obtains over 96.5% accurate results with an average of 30-ms running time. The proposed scheme is also validated on an open dataset DAIR (Dataset for AIR Handwriting), where it achieves 95.5% mean accuracy with 24.3-ms recognition time per gesture. The developed approach is compared with benchmark schemes to justify its accuracy and speed.