Since the Transformer model's development, its influence on diverse machine learning fields has been substantial and multifaceted. Time series prediction's advancement has also been fueled by the proliferation of Transformer models, resulting in a range of differentiated variants. Multi-head attention mechanisms in Transformer models amplify the effectiveness of attention mechanisms used for feature extraction. Nevertheless, multi-head attention fundamentally represents a straightforward overlay of identical attention mechanisms, thereby failing to ensure the model's capacity to discern diverse features. Conversely, multi-head attention mechanisms can introduce substantial redundancy in the information processed, resulting in wasted computational resources. This paper introduces a hierarchical attention mechanism to the Transformer, for the first time. This mechanism is designed to better capture information from multiple perspectives, thus improving feature diversity. The proposed mechanism overcomes the drawbacks of traditional multi-head attention mechanisms, which struggle with insufficient information diversity and lack of interaction among different heads. Graph networks are further employed to aggregate global features, which helps to mitigate inductive bias. Our final experiments on four benchmark datasets reveal that the proposed model exhibits superior performance compared to the baseline model in various metrics.
Livestock breeding benefits significantly from insights gleaned from changes in pig behavior, and the automated recognition of pig behavior is essential for boosting animal welfare. However, a significant portion of approaches to identifying pig behaviors are contingent upon human observation and the use of deep learning. Though human observation often demands a considerable investment of time and effort, deep learning models, despite their large parameter sets, may nonetheless present challenges concerning slow training times and efficiency. This paper proposes a novel, two-stream pig behavior recognition methodology, leveraging deep mutual learning, to address the identified issues. The proposed model's structure involves two networks that learn from each other, which use the red-green-blue color model and flow streams. Each branch, in addition, features two student networks that learn cooperatively, producing detailed and rich visual or motion attributes, leading to better detection of pig behaviors. To further refine pig behavior identification, the RGB and flow branch results are weighted and integrated. Experimental validations unequivocally highlight the prowess of the proposed model, achieving top-tier recognition accuracy of 96.52%, exceeding other models by a remarkable 2.71 percentage points.
For improved maintenance practices concerning bridge expansion joints, the utilization of IoT (Internet of Things) technology is highly significant. predictive toxicology To pinpoint faults in bridge expansion joints, a high-efficiency, low-power end-to-cloud coordinated monitoring system leverages acoustic signals. Recognizing the dearth of genuine data on bridge expansion joint failures, a data collection platform for simulating expansion joint damage, with meticulous annotation, is established. This paper introduces a progressive two-tiered classifier combining template matching, leveraging AMPD (Automatic Peak Detection), and deep learning algorithms based on VMD (Variational Mode Decomposition) for denoising, all while efficiently utilizing edge and cloud computing. To assess the efficacy of the two-level algorithm, simulation-based datasets were used. The first-level edge-end template matching algorithm achieved a remarkable fault detection rate of 933%, while the second-level cloud-based deep learning algorithm attained a classification accuracy of 984%. The preceding results support the claim that the proposed system in this paper has demonstrated efficient performance in monitoring the health of expansion joints.
Rapid updates to traffic signs necessitate substantial manpower and material resources for image acquisition and labeling, hindering the generation of ample training data crucial for high-precision recognition. Transjugular liver biopsy For the purpose of resolving this issue, a new traffic sign recognition approach, based on few-shot object discovery (FSOD), is put forward. The original model's backbone network is modified by this method, incorporating dropout to enhance detection accuracy and mitigate overfitting. Furthermore, a refined RPN (region proposal network), incorporating an enhanced attention mechanism, is introduced to produce more precise bounding boxes for target objects by selectively highlighting specific characteristics. The final component for multi-scale feature extraction is the FPN (feature pyramid network), which integrates high-semantic, low-resolution feature maps with high-resolution, but less semantically rich feature maps, leading to a more precise detection outcome. The improved algorithm performs 427% better on the 5-way 3-shot task and 164% better on the 5-way 5-shot task when contrasted with the baseline model. We perform an application of the model's structure using the PASCAL VOC dataset. According to the results, this method exhibits a clear advantage over a selection of current few-shot object detection algorithms.
The cold atom absolute gravity sensor (CAGS), leveraging cold atom interferometry, stands out as a cutting-edge high-precision absolute gravity sensor, indispensable for advancements in scientific research and industrial technologies. A significant obstacle to the real-world implementation of CAGS on mobile platforms is the combination of its large size, considerable weight, and high energy consumption. With cold atom chips, a reduction in the weight, size, and complexity of CAGS is achievable. Using the basic principles of atom chips as our point of departure, this review constructs a comprehensive progression toward related technologies. selleck chemical Micro-magnetic traps and micro magneto-optical traps, in conjunction with material selection procedures, fabrication processes, and packaging strategies, were amongst the discussed related technologies. This paper gives a detailed account of the current evolution of cold atom chip technology, highlighting various implementations and featuring discussions of practical applications in CAGS systems arising from atom chips. We conclude by listing several of the challenges and possible future research directions in this field.
The presence of dust or condensed water in harsh outdoor environments, or in human breath with high humidity, is a primary reason for erroneous results when using Micro Electro-Mechanical System (MEMS) gas sensors. This paper proposes a novel MEMS gas sensor packaging, characterized by a self-anchoring integration of a hydrophobic PTFE filter within the gas sensor's upper cover. This method diverges significantly from the existing procedure of external pasting. This research successfully demonstrates the functionality of the proposed packaging mechanism. Analysis of the test results shows that the innovative packaging incorporating a PTFE filter decreased the sensor's average response to humidity levels ranging from 75% to 95% RH by 606% in comparison to the packaging without the PTFE filter. Furthermore, the packaging demonstrated its reliability through successful completion of the High-Accelerated Temperature and Humidity Stress (HAST) test. A similar sensing system integrated within the proposed packaging with a PTFE filter could further facilitate the application of breath screening for conditions linked to exhalation, including coronavirus disease 2019 (COVID-19).
Congestion is unavoidable for millions of commuters, a part of their everyday routines. The key to mitigating traffic congestion lies in the careful application of effective transportation planning, design, and management techniques. Accurate traffic data are crucial for making well-informed decisions. To this end, operational bodies install permanent and often temporary detectors on public roads for calculating the movement of cars. This traffic flow measurement is the cornerstone for estimating demand across the network. Nevertheless, detectors fixed in place are distributed thinly across the road system, failing to encompass the entire network, and temporary detectors are thinly distributed over time, often yielding only a few days of data every couple of years. Considering the current situation, previous research proposed that public transit bus fleets could be transformed into surveillance assets if outfitted with additional sensors. The robustness and precision of this strategy were confirmed by the manual analysis of visual data captured by cameras installed on the transit buses. By leveraging the existing perception and localization sensors on these vehicles, we propose to operationalize this traffic surveillance methodology for practical use cases in this paper. We describe an automatic vehicle counting system that is based on vision, using video data from cameras positioned on transit buses. In a state-of-the-art fashion, a 2D deep learning model identifies objects, processing each frame individually. Thereafter, tracked objects utilize the frequently employed SORT method. The proposed counting mechanism reinterprets tracking results to provide vehicle totals and their bird's-eye-view paths in the real world. Video imagery collected from active transit buses over multiple hours allowed us to demonstrate our system's ability to pinpoint and track vehicles, discern parked vehicles from those in traffic, and count vehicles in both directions. Under diverse weather conditions, the proposed method's effectiveness in accurately counting vehicles is demonstrated through an exhaustive ablation study and analysis.
City residents endure the ongoing ramifications of light pollution. Nocturnal light pollution significantly disrupts the human circadian rhythm. Accurate measurement of light pollution levels across urban areas is critical for targeted reductions where appropriate.