Skeleton-based human activity recognition is a key research topic in the fields of deep learning and computer vision. However, existing approaches are less effective at capturing short-term sub-action information at different granularity levels and long-term motion correlations, which affect recognition accuracy. To overcome these challenges, an innovative multi-grained temporal clip transformer (MTC-Former) is proposed. Firstly, based on the transformer backbone, a multi-grained temporal clip attention (MTCA) module with multi-branch architecture is proposed to capture the characteristics of short-term sub-action features. Secondly, an innovative multi-scale spatial–temporal feature interaction module is proposed to jointly learn sub-action dependencies and facilitate skeletal motion interactions, where long-range motion patterns are embedded to enhance correlation modeling. Experiments were conducted on three datasets, including NTU RGB+D, NTU RGB+D 120, and InHARD, and achieved state-of-the-art Top-1 recognition accuracy, demonstrating the superiority of the proposed MTC-Former.

Distributed photovoltaic (DPV) power sites in industrial parks are characterized by dispersed layouts, practical fault detection environments, and high safety requirements. Conventional manual DPV O&M systems using handheld sensors are inefficient, expensive, and struggle with fault detection due to sparse industrial data and uni-modal information limitations. To this, this paper proposes an innovative advanced algorithm for DPV fault detection in industrial parks, utilizing a new sparse industrial dataset, "SolarPark," collected via multi-modal UAVs and annotated through a multi-expert process with uncertainty scoring. By fusing the Convolutional Block Attention Module (CBAM), Bidirectional Feature Pyramid Network (BiFPN), Ghost modules, the algorithm enhances attention to critical photovoltaic fault-related channel information, strengthens multi-scale photovoltaic fault feature fusion, and achieves lightweight efficiency. Combined with multi-modal UAV videos, the proposed industrial DPV fault detection algorithm achieves a precision of 95.4%, effectively ensuring the efficiency of DPV power sites in data-scarce industrial scenarios. Extensive experiments on the developed cloud platform confirm the proposed algorithm’s efficient, cost-effective, and easy to deploy for aerial inspections of DPV O&M systems.

In video human action-recognition tasks, motion tempo describes the dynamic patterns and temporal scales of human motion. Different categories of actions are typically composed of sub-actions with varying motion tempos. Effectively capturing sub-actions with different motion tempos and distinguishing category-specific sub-actions are crucial for improving action-recognition performance. Convolutional Neural Network (CNN)-based methods attempted to address this challenge, by embedding feedforward attention modules to enhance the action’s dynamic representation learning. However, feedforward attention modules rely only on local information from low-level features, lacking contextual information to generate attention weights. Therefore, we propose a Sub-action Motion information Enhancement Network (SMEN) based on motion-tempo learning and feedback attention, which consists of the Multi-Granularity Adaptive Fusion Module (MgAFM) and Feedback Attention-Guided Module (FAGM). MgAFM enhances the model’s ability to capture crucial sub-action intrinsic information by extracting and adaptively fusing motion dynamic features at different granularities. FAGM leverages high-level features that contain contextual information in a feedback manner to guide low-level features in generating attention weights, enhancing the model’s ability to extract more discriminative spatio-temporal and channel-wise features. Experiments are conducted on three datasets, and the proposed SMEN achieves top-1 accuracies of 52.4%, 63.3% on the Something-Something V1 and V2 datasets, and 76.9% on the Kinetics-400 dataset. Ablation studies, evaluations, and visualizations demonstrate that the proposed SMEN is effective for sub-action motion tempo and representation learning, and outperforms compared methods for video action recognition.