Few-shot fine-grained image classification remains a challenging task due to the subtle inter-class variations and the scarcity of labeled data. Existing few-shot fine-grained methods often struggle to generalize effectively under low-data regimes and fail to discriminate between visually similar samples. To address these challenges, we propose DistillCvT, a two-stage framework to enhance model generalization and feature discrimination through self-supervised learning and self-distillation within prototypical network structure. Convolutional Vision Transformer (CvT) is employed as the feature extractor, which integrates convolutional locality with transformer-based global modeling to capture fine-grained details while preserving global context. DistillCvT mitigates these issues through a two-stage strategy. In the first stage, the model is trained using a prototypical distance-based classification loss, where class probabilities are computed from the distances between query samples and class prototypes, complemented by standard data augmentation to mitigate data scarcity and learn discriminative class representations. In the second stage, an identical student model is trained with a self-supervised jigsaw task to enhance feature discrimination and robustness by aligning support samples with their augmented query counterparts. Simultaneously, the teacher model from stage one guides it through logit-based self-distillation to further improve generalization. Experiments on CUB-200-2011, Stanford Dogs, and Stanford Cars demonstrate that DistillCvT consistently improves performance and outperforms recent state-of-the-art few-shot fine-grained image classification methods.

The learning-based, passive, device-free wireless human activity recognition (WHAR) systems still face significant challenges, especially in real-world deployments. Environmental differences and domain diversities cause signals collected in the source domain to have a different distribution from those in the target domain, and this affects the accuracy. To achieve domain generalization (DG), a multi-scale linear attention vision transformer (ViT) based feature extractor and domain adversarial learning with Wasserstein distance are proposed. By aligning both marginal and conditional distributions across different source domains, the adversarial learning reduces the differences between trained and unseen domains. As a result, the extracted features become domain-invariant in the latent space, ensuring accuracy is preserved in new or unseen domains. Extensive evaluations using commercial IEEE 802.11ac routers with human activity data collected over different days, environments, human subjects, and obstacle configurations show that the proposed Wi-ViTAL achieves 97.57% average accuracy for five-label classification and more than 76% for eight-label classification in unseen domains. Wi-ViTAL also demonstrates an overall DG improvement compared to other recent benchmarks.

Over the past decade, deep learning has made significant advancements in image classification. However, these models struggle with data scarcity and distribution shifts, commonly referred to as the few-shot image classification (FSIC) problem. FSIC aims to recognize novel classes using only a limited number of labeled samples, posing challenges for conventional deep learning models that rely on large datasets for optimal performance. This paper provides a comprehensive review of FSIC methodologies, categorizing them into five main approaches: meta-learning, transfer learning, data augmentation, attribute-related, and vision-language foundation model adaptation. Meta-learning approaches are further classified into metric-based, model-based, and optimization-based methods, while transfer learning approaches are divided into hybrid and non-hybrid methods. Vision-language foundation model adaptation approaches are grouped into few-shot parameter tuning, dynamic or unsupervised tuning, and training-free adaptation methods. Beyond general FSIC, this paper also explores specialized FSIC methods in fine-grained classification, cross-domain classification, and class-incremental learning. Additionally, it reviews commonly used few-shot image datasets and compares the performance of representative methods through experimental results. Practical applications of FSIC across various domains are also discussed, highlighting its potential to address real-world challenges. Finally, the research trends of FSIC are identified, offering insights into the state-of-the-art FSIC methods and guiding future advancements in this field.

Typically, hyperspectral image suffers from redundant information, data scarcity, and class imbalance problems. This letter proposes a hyperspectral image classification framework named a two-stream residual separable convolution (2SRS) network that aims to mitigate these problems. Principal component analysis (PCA) is first employed to reduce the spectral dimension of the hyperspectral image. Subsequently, the data scarcity and class imbalance problems are overcome via spatial and spectral data augmentations. A novel spectral data creation from image patches is proposed. The augmented samples are fed into the proposed 2SRS network for hyperspectral image classification. We evaluated the proposed method on three benchmark datasets, namely, 1) Indian Pines (IP); 2) Pavia University; and 3) Salinas Scene (SA). The proposed method achieved state-of-the-art performance in terms of overall accuracy (OA), average accuracy (AA), and Kappa coefficient (Kappa) for both 30% and 10% training set ratios.

In this project, an experimental study on disease detection and nutrient extraction of plantation such as oil palm will be carried out using a drone based hyperspectral imaging system developed by Centre for Remote Sensing and Surveillance Technologies (CRSST), Multimedia University (MMU), Malaysia. The major advantages of this system are light weight (approximately 250g for sensor) and larger number of band selection (from 500nm to 900nm) compared to a conventional multispectral camera. A customised multirotor drone has been designed and developed for longer endurance operation and to carry the non-standard payload i.e. hyperspectral camera. Preliminary testing has been performed in laboratory and oil palm plantation to verify the developed system. Initial results show that the hyperspectral data are suitable to be used for differentiation of the healthiness level of the oil palm plantation.

In this paper, we propose a hand tracking method which was inspired by the notion of the four dukkha: birth, aging, sickness and death (BASD) in Buddhism. Based on this philosophy, we formalize the hand tracking problem in the BASD framework, and apply it to hand track hand gestures in isolated sign language videos. The proposed BASD method is a novel nature-inspired computational intelligence method which is able to handle complex real-world tracking problem . The proposed BASD framework operates in a manner similar to a standard state-space model, but maintains multiple hypotheses and integrates hypothesis update and propagation mechanisms that resemble the effect of BASD. The survival of the hypothesis relies upon the strength, aging and sickness of existing hypotheses, and new hypotheses are birthed by the fittest pairs of parent hypotheses. These properties resolve the sample impoverishment problem of the particle filter. The estimated hand trajectories show promising results for the American sign language .