Search

Object detection, as a crucial aspect of computer vision, plays a vital role in traffic management, emergency response, autonomous vehicles, and smart cities. Despite the significant advancements in object detection, detecting small objects in images captured by high-altitude cameras remains challenging, due to factors such as object size, distance from the camera, varied shapes, and cluttered backgrounds. To address these challenges, we propose small object detection YOLOv8 (SOD-YOLOv8), a novel model specifically designed for scenarios involving numerous small objects. Inspired by efficient generalized feature pyramid networks (GFPNs), we enhance multi-path fusion within YOLOv8 to integrate features across different levels, preserving details from shallower layers and improving small object detection accuracy. Additionally, we introduce a fourth detection layer to effectively utilize high-resolution spatial information. The efficient multi-scale attention module (EMA) in the C2f-EMA module further enhances feature extraction by redistributing weights and prioritizing relevant features. We introduce powerful-IoU (PIoU) as a replacement for CIoU, focusing on moderate quality anchor boxes and adding a penalty based on differences between predicted and ground truth bounding box corners. This approach simplifies calculations, speeds up convergence, and enhances detection accuracy. SOD-YOLOv8 significantly improves small object detection, surpassing widely used models across various metrics, without substantially increasing the computational cost or latency compared to YOLOv8s. Specifically, it increased recall from 40.1% to 43.9%, precision from 51.2% to 53.9%, mAP0.5 from 40.6% to 45.1%, and mAP0.5:0.95 from 24% to 26.6%. Furthermore, experiments conducted in dynamic real-world traffic scenes illustrated SOD-YOLOv8's significant enhancements across diverse environmental conditions, highlighting its reliability and effective object detection capabilities in challenging scenarios. [ABSTRACT FROM AUTHOR]

The integration of data‐driven models such as neural networks for high‐consequence decision making has been largely hindered by their lack of predictive power away from training data and their inability to quantify uncertainties often prevalent in engineering applications. This article presents an ensembling method with function‐space regularization, which allows to integrate prior information about the function of interest, thus improving generalization performance, while enabling quantification of aleatory and epistemic uncertainties. This framework is applied to build a probabilistic ambulance travel time predictor, leveraging historical ambulance data provided by the Fire Department of New York City. Results show that the integration of a non‐Gaussian likelihood and prior information from a road network analysis yields appropriate probabilistic predictions of travel times, which could be further leveraged for emergency medical service (EMS) decision making. [ABSTRACT FROM AUTHOR]

The adaptive parametric identification of deteriorating and nondeteriorating nonlinear hysteretic phenomena is considered using a generalization of Masing model based on the observed memory behavior of distributed element models. The model permits a parametric identification to be performed using nonlinear optimization techniques for arbitrary response time histories. A changing objective function, defined as the normalized force estimation error over a shifting window of recent data, is employed so that classic nonlinear optimization techniques can be used for the adaptive identification problem. A variation of the steepest descent method is used with significant modifications. To achieve the best performance for any given problem, a set of a priori numeric tests are suggested to design the identification scheme. The design identification scheme exhibits a very good performance in identifying the correct values of the parameters and is rather robust in dealing with noise. The proposed approach has applications to adaptive identification of much wider types of nonlinear rate-dependent hysteretic behavior. Also, the set of guidelines proposed by the authors is a contribution toward having more effective autonomous identification schemes, using minimal information about the model and input. DOI: 10.1061/(ASCE)0733-9399(2008)134:6(482) CE Database subject headings: Parameters; Deterioration; Hysteresis.

The modeling of hysteretic behavior is of significant importance in several areas, including civil engineering and mechanics. This paper focuses on finding a method for modeling hysteretic behavior that is simple and efficient in terms of capturing the response and retaining the memory, if any, and at the same time is proper for use in physically meaningful modeling and identification of the system with few parameters. A distributed-element model (DEM) capable of capturing deterioration is used as a starting point, and its characteristics are studied, with a particular focus on the way memory is stored in the model. It is observed that keeping track of the response at a few of the past extremes of input displacement, called the Sequence of Dominant Alternating Extremes, is enough for representing the effect of history. The relation of this behavior to a generalized Masing model is studied. A set of rules is proposed which is a generalization of the Masing rules and can capture the deteriorating (or nondeteriorating) response of DEMs with any distribution of element yield displacement thresholds to any arbitrary loading. The presented formulation provides a framework for efficient modeling and identification of dynamic models of very different characteristics with only a few physically meaningful parameters. DOI: 10.1061/(ASCE)0733-9399(2007)133:5(495) CE Database subject headings: Hysteresis; Constitutive models; Deterioration; Identification; Ductility; Nonlinear response; Loading history.

Based on the basic formulation developed in a companion paper, the writers now present the application of an artificial neural network approach to designing streamlined network models to simulate and identify the nonlinear dynamic response of single-degree-of-freedom oscillators using the restoring force-state mapping interpretation. The neural networks which use sigmoidal activation functions are shown to be highly robust in modeling a wide variety of commonly observed nonlinear structural dynamic response behaviors. By streamlining the networks, individual network model parameters take on physically or geometrically interpretable meaning, and hence, the network initialization can be achieved through an engineered approach rather than through less physically meaningful numerical initialization schemes. Although not proven in general, examples show that by starting with a more meaningful initial design, identification convergence is improved, and the final identified model parameters are seen to have a more physical meaning. A set of model architecture prototypes is developed to capture commonly observed nonlinear response behaviors. CE Database subject headings: Neural networks; Dynamic response; Models; Parameters; Design.

This paper addresses the modeling problem of nonlinear and hysteretic dynamic behaviors through a constructive modeling approach which exploits existing mathematical concepts in artificial neural network modeling. In contrast with many neural network applications, which often result in large and complex 'black-box' models, here, the writers strive to produce phenomenologically accurate model behavior starting with network architecture of manageable/small sizes. This affords the potential of creating relationships between model parameter values and observed phenomenological behaviors. Here a linear sum of basis functions is used in modeling nonlinear hysteretic restoring forces. In particular, nonlinear sigmoidal activation functions are chosen as the core building block for their robustness in approximating arbitrary functions. The appropriateness and effectiveness of this set of basis function in modeling a wide variety of nonlinear dynamic behaviors observed in structural mechanics are depicted from an algebraic and geometric perspective. CE Database subject headings: Neural networks; Mathematics; Geometry; Models; Hysteresis.

Current practice in buckling design is based on the ultimate strength of a member such that excessive lateral deflections do not occur. This ultimate strength is often dictated by empirically derived formulae which correspond to postulated element imperfections including residual stresses. In this paper, the postbuckling behavior of columns is considered, and the limit state is considered to be total collapse. The load versus end displacement relationship is derived for a simplified elastic-perfectly plastic column model. Using this curve, the energy absorbed and released in going from one load/deflection state to another can be easily quantified. With this, the energy absorption capacity of an already loaded column is considered to be the critical performance metric. This 'energy capacity criterion' is then used to develop alternative design curves for different levels of energy absorption capacity. DOI: 10.1061/(ASCE)0733-9399(2006) 132:6(594) CE Database subject headings: Columns; Design; Deflection; Collapse; Ultimate strength; Buckling.