Fine-Grained Image Classification (FGIC) remains a complex task in computer vision, as it requires models to distinguish between categories with subtle localized visual differences. Well-studied CNN-based models, whil...
详细信息
A conceptual ontological model of Digital Crime has been developed, consisting of five non-empty classes. Identification and classification were carried out according to the experience of domestic and foreign experts....
详细信息
The world is currently experiencing population explosion, and it has been forecast that the population would reach nine billion by the year 2050. This rapid population growth is accompanied with new challenges, includ...
详细信息
Small object detection aims to localize and classify small objects within images. With recent advances in large-scale vision-language pretraining, finetuning pretrained object detection models has emerged as a promisi...
详细信息
ISBN:
(数字)9798350368741
ISBN:
(纸本)9798350368758
Small object detection aims to localize and classify small objects within images. With recent advances in large-scale vision-language pretraining, finetuning pretrained object detection models has emerged as a promising approach. However, finetuning large models is computationally and memory expensive. To address this issue, this paper introduces multi-point positional insertion (MPI) tuning, a parameter-efficient finetuning (PEFT) method for small object detection. Specifically, MPI incorporates multiple positional embeddings into a frozen pretrained model, enabling the efficient detection of small objects by providing precise positional information to latent features. Through experiments, we demonstrated the effectiveness of the proposed method on the SODA-D dataset. MPI performed comparably to conventional PEFT methods, including CoOp and VPT, while significantly reducing the number of parameters that need to be tuned.
Charging a significant number of electric vehicles (EVs) can pose challenges for power systems, and a viable solution is smart charging. Reliable smart charging can be achieved by forecasting the required energy and p...
详细信息
ISBN:
(数字)9798331511333
ISBN:
(纸本)9798331511340
Charging a significant number of electric vehicles (EVs) can pose challenges for power systems, and a viable solution is smart charging. Reliable smart charging can be achieved by forecasting the required energy and plug-in duration of a charging event. However, despite the availability of data from public charging stations, the lack of sufficient real data for residential EV charging has consistently been a barrier, particularly when using deep learning methods. Moreover, enhancing the accuracy of forecasting residential EV charging behavior is crucial due to its significant impact on power systems. To address these challenges, this paper proposes an intelligent deep learning-based framework to forecast the required energy and plug-in duration of residential EV charging events by leveraging transfer learning. Additionally, the proposed framework utilizes an auxiliary deep learning model that forecasts charging duration to improve the accuracy of required energy forecasting. Various case studies have been conducted, and the results show that the proposed framework significantly improves the accuracy of residential charging behavior forecasting.
Robots playing soccer often rely on hard-coded behaviors that struggle to generalize when the game environment change. In this paper, we propose a temporal logic based approach that allows robots’ behaviors and goals...
详细信息
mobile robots play an enormous role in different fields of daily life applications including military, safety, and logistic multi-tasking capabilities. A new approach is introduced to the market which is the 3 Mecanum...
详细信息
This paper addresses model predictive control of a class of linear systems subject to additive stochastic disturbances and constraints. The underlying stochastic optimal control problem combines inverse cumulative dis...
详细信息
This paper addresses model predictive control of a class of linear systems subject to additive stochastic disturbances and constraints. The underlying stochastic optimal control problem combines inverse cumulative distribution functions with ellipsoid-in-polyhedron formulations to reduce the conservatism induced by constraint satisfaction. By use of terminal constraints and time-varying weights within the cost functional, the presented control scheme satisfies criteria for mean-square stability and can be adapted to reference-tracking problems for arbitrary reference signals.
Demand response is expected to play a fundamental role in providing flexibility for balancing operations to the grid. On the other hand, the fast electrification of the transportation sector calls for new solutions to...
Demand response is expected to play a fundamental role in providing flexibility for balancing operations to the grid. On the other hand, the fast electrification of the transportation sector calls for new solutions to enforce safe and reliable grid operation. Here we consider an electric vehicle charging station that participates in demand response programs. The demand response program asks for a change of the charging station load profile in exchange for a monetary reward. A stochastic receding horizon scheme that exploits the charging flexibility is then designed to optimally coordinate vehicle charging. Numerical simulations show that the proposed approach ensures substantial cost reduction compared to simpler benchmarks while maintaining the computation time feasible for real-world applications.
This paper introduces the Transient Predictor and describes how it can be used to estimate the Multistep Predictor, which can be applied to applications such as Data-Driven Predictive control (DDPC). The Transient Pre...
详细信息
ISBN:
(数字)9798350316339
ISBN:
(纸本)9798350316346
This paper introduces the Transient Predictor and describes how it can be used to estimate the Multistep Predictor, which can be applied to applications such as Data-Driven Predictive control (DDPC). The Transient Predictor has two desirable traits that differentiate it from other methods for estimating the Multistep Predictor, such as the standard Subspace Predictor method: 1) Causality-the Transient Predictor asserts a causal relationship between future inputs and future outputs; and 2) Bias-the Transient Predictor is a consistent predictor of future outputs. This paper provides an easy-toimplement algorithm for estimating the Transient Predictor and in turn the Multistep Predictor, and demonstrates its efficacy for DDPC. In experiments, we find that the Transient Predictorbased DDPC performs remarkably well with small lead-in data lengths, indicating that it is well-suited for tasks in which large amounts of data are not available. In addition, the Transient Predictor is not afflicted by the same bias as subspace-based methods when data is gathered in closed loop.
暂无评论