arXiv subjects

Thomas Motz

Publications and source records attributed to Thomas Motz.

7 recordsLinked to original sources

arXiv2024

OCDet: Object Center Detection via Bounding Box-Aware Heatmap Prediction on Edge Devices with NPUs

Real-time object localization on edge devices is fundamental for numerous applications, ranging from surveillance to industrial automation. Traditional frameworks, such as object detection, segmentation, and keypoint detection, struggle in resource-constrained environments, often resulting in substantial target omissions. To address these challenges, we introduce OCDet, a lightweight Object Center Detection framework optimized for edge devices with NPUs. OCDet predicts heatmaps representing object center probabilities and extracts center points through peak identification. Unlike prior methods using fixed Gaussian distribution, we introduce Generalized Centerness (GC) to generate ground truth heatmaps from bounding box annotations, providing finer spatial details without additional manual labeling. Built on NPU-friendly Semantic FPN with MobileNetV4 backbones, OCDet models are trained by our Balanced Continuous Focal Loss (BCFL), which alleviates data imbalance and focuses training on hard negative examples for probability regression tasks. Leveraging the novel Center Alignment Score (CAS) with Hungarian matching, we demonstrate that OCDet consistently outperforms YOLO11 in object center detection, achieving up to 23% higher CAS while requiring 42% fewer parameters, 34% less computation, and 64% lower NPU latency. When compared to keypoint detection frameworks, OCDet achieves substantial CAS improvements up to 186% using identical models. By integrating GC, BCFL, and CAS, OCDet establishes a new paradigm for efficient and robust object center detection on edge devices with NPUs. The code is released at https://github.com/chen-xin-94/ocdet.

cs.CV↗

arXiv2022

Deep Learning-Based Position Detection for Hydraulic Cylinders Using Scattering Parameters

Position detection of hydraulic cylinder pistons is crucial for numerous industrial automation applications. A typical traditional method is to excite electromagnetic waves in the cylinder structure and analytically solve the piston position based on the scattering parameters measured by a sensor. The core of this approach is a physical model that outlines the relationship between the measured scattering parameters and the targeted piston position. However, this physical model has shortcomings in accuracy and adaptability, especially in extreme conditions. To address these limitations, we propose machine learning and deep learning-based methods to learn the relationship directly in a data-driven manner. As a result, all deep learning models in this paper consistently outperform the physical one by a large margin. We further deliberate on the choice of models based on domain knowledge and provide in-depth analyses combining model performance with real-world physical characteristics. Specifically, we use Convolutional Neural Network (CNN) to discover local interactions of input among adjacent frequencies, apply Complex-Valued Neural Network (CVNN) to exploit the complex-valued nature of electromagnetic scattering parameters, and introduce a novel technique named Frequency Encoding to add weighted frequency information to the model input. The combination of these techniques results in our best-performing model, a complex-valued CNN with Frequency Encoding, which exhibits substantial improvement in accuracy with an error reduction of 1/12 compared to the traditional physical model.

eess.SP↗

arXiv2021

Explainable Online Validation of Machine Learning Models for Practical Applications

We present a reformulation of the regression and classification, which aims to validate the result of a machine learning algorithm. Our reformulation simplifies the original problem and validates the result of the machine learning algorithm using the training data. Since the validation of machine learning algorithms must always be explainable, we perform our experiments with the kNN algorithm as well as with an algorithm based on conditional probabilities, which is proposed in this work. For the evaluation of our approach, three publicly available data sets were used and three classification and two regression problems were evaluated. The presented algorithm based on conditional probabilities is also online capable and requires only a fraction of memory compared to the kNN algorithm.

cs.LG↗

arXiv2018

Rectification of heat currents across nonlinear quantum chains: A versatile approach beyond weak thermal contact

Within the emerging field of quantum thermodynamics the issues of heat transfer and heat rectification are basic ingredients for the understanding and design of heat engines or refrigerators at nanoscales. Here, a consistent and versatile approach for mesoscopic devices operating with continuous degrees of freedom is developed valid from low up to strong system-reservoir couplings and over the whole temperature range. It allows to cover weak to moderate nonlinearities and is applicable to various scenarios including the presence of disorder and external time-dependent fields. As a particular application coherent one-dimensional chains of anharmonic oscillators terminated by thermal reservoirs are analyzed with particular focus on rectification. The efficiency of the method opens a door to treat also rather long chains and extensions to higher dimensions and geometries.

cond-mat.stat-mech↗

arXiv2017

Currents and fluctuations of quantum heat transport in harmonic chains

Heat transport in open quantum systems is particularly susceptible to the modeling of system-reservoir interactions. It thus requires to consistently treat the coupling between a quantum system and its environment. While perturbative approaches are successfully used in fields like quantum optics and quantum information, they reveal deficiencies, typically in the context of thermodynamics, when it is essential to respect additional criteria such as fluctuation-dissipation theorems. We use a non-perturbative approach for quantum dissipative dynamics based on a stochastic Liouville-von Neumann equation to provide a very general and extremely efficient formalism for heat currents and its correlations in open harmonic chains. Specific results are derived not only for first but also for second order moments which requires to account for both real and imaginary parts of bath-bath correlation functions. Spatiotemporal patterns are compared with weak coupling calculations. The regime of stronger system-reservoir couplings gives rise to an intimate interplay between reservoir fluctuations and heat transfer far from equilibrium.

cond-mat.stat-mech↗

arXiv2016

Thermodynamic deficiencies of some simple Lindblad operators

Master equations of Lindblad type have attained prominent status in the fields of quantum optics and quantum information since they are guaranteed to satisfy fundamental notions of quantum dynamics such as complete positivity. When Lindblad operators are used to describe thermal reservoirs in contact with an open quantum system, the fundamental laws of thermodynamics and the fluctuation-dissipation theorem provide additional mandatory criteria. We show several examples of innocent-looking Lindblad operators which have questionable properties in this regard. Compatibility criteria between Hamiltonian and Lindblad terms as well as consequences of their violation are discussed. An alternative stochastic approach to dissipative quantum dynamics is outlined and illustrated through a harmonic-chain model for which the approach of local Lindblad operators fails.

cond-mat.stat-mech↗

arXiv2014

Optimizing the Performance of the Entropic Splitter for Particle Separation

Recently, it has been shown that entropy can be used to sort Brownian particles according to their size. In particular, a combination of a static and a time-dependent force applied on differently sized particles which are confined in an asymmetric periodic structure can be used to separate them efficiently, by forcing them to move in opposite directions. In this paper, we investigate the optimization of the performance of the 'entropic splitter'. Specifically, the splitting mechanism and how it depends on the geometry of the channel, and the frequency and strength of the periodic forcing is analyzed. Using numerical simulations, we demonstrate that a very efficient and fast separation with a practically 100% purity can be achieved by a proper optimization of the control variables. The results of this work could be useful for a more efficient separation of dispersed phases such as DNA fragments or colloids dependent on their size.

physics.chem-ph↗