Tooth instance segmentation from cone-beam CT images through point-based detection and Gaussian disentanglement

Multimedia Tools and Applications - Individual tooth segmentation and identification from cone-beam computed tomography images are preoperative prerequisites for orthodontic treatments. Instance...     

Design and analysis of an infrared range sensor system for floor-state estimation

Most blind and vision-impaired people use a white cane or a guide dog to aid their walking. The conventional white cane helps the user to perceive the environment simply by identifying the presence of nearby obstacles. In the present study, three infrared range sensors were used to identify the floor state. The sensor system, affixed to the user’s belt, does not require swinging motion or any other movement by the user. The three sensors are configured radially, and the design parameters are selected based on the sensor characteristics. The main difference from the conventional range sensor system for the blind is that the infrared range sensor system identifies three discrete states (even surface, ascending stair, descending stair). The condition for the state transition was derived from a sensor system model. Additionally, the effects of sensor system movement were verified by covariance analysis.     

Modification of hydrogenated Bisphenol A epoxy adhesives using nanomaterials

Color-stable hydrogenated Bisphenol A (HBA) epoxy adhesives, containing organic-inorganic hybrid nanomaterials, were prepared and their properties investigated. Poly(propylene glycol)bis(2-aminopropyl ether) (D230) was used as the room temperature curing agent, and functional organic-inorganic hybrid nanomaterials, to tailor the adhesives, were prepared by a sol-gel reaction of 3-glycidoxypropyltrimethoxysilane and tetraethoxysilane. The commercial polyhedral oligomeric silsesquioxanes (POSS) having epoxy functional groups were also used. The concentration dependence of different nanomaterials, containing epoxy functional group for HBA/D230 adhesives, was studied. The tensile strength increased with the addition of nanomaterials having glycidyl epoxy group; however, the dependence varied with the size, the number of functional groups, and the amount of the addition. HBA/D230 adhesives containing different amounts of nanomaterials, whose compositions are similar to that of granite, were applied to the Korean granite and the results were compared with those obtained by using commercial adhesives, which have the problem of significant color change and high viscosity. The mechanical properties of HBA/D230 adhesives, containing POSS having glycidyl epoxy group, are found to be similar to those of commercial adhesives. Besides, it has low viscosity and long-term color stability.     

Reduced graphene oxide-reinforced unsaturated polyester resin nanocomposites with improved thermal stability,mechanical modulus,and electromagnetic interference shielding performance

This article aims to investigate the impact of reduced graphene oxide (RGO) nanofillers on the curing kinetics, thermal stability, mechanical modulus, electrical conductivity, and EMI shielding effectiveness of unsaturated polyester resin (UPR). The curing rates of UPR/styrene (60/40 by wt%) mixtures with small amounts of RGO (0.1–0.3 wt%) exhibit slight delays owing to the barrier and scavenger roles of 2-dimensional RGO sheets. Nonetheless, it is observed that within the cured nanocomposites, RGOs are effectively dispersed and firmly bonded to the UPR matrix at interfaces through hydrogen bonding and π-π interactions. Consequently, the nanocomposites display heightened thermal decomposition temperatures and increased residue at 800°C with higher RGO loading content. The addition of RGO notably improves the elastic storage modulus and increases the temperature associated with glass transition-related relaxation. The electrical percolation threshold is attained at a specific RGO loading between 0.2 and 0.3 wt%. Thus, the nanocomposite with 0.3 wt% RGO is characterized to have an electrical conductivity of 1.9 × 10⁻⁶ S/cm and an EMI shielding effectiveness of ~9 dB at 8 GHz, for a thickness of 1 mm.     

High temperature hardness of tungsten carbide

The effect of crystallographic orientation and test temperature on hardness of WC single crystals was investigated along with the hot hardness of poly crystalline tungsten carbide. Also investigated was the effect of carbide grain size and the amount of binder phase on the hot hardness of some cemented tungsten carbides. The hot hardness of single crystal WC on all major crystallographic orientations evaluated decreases very rapidly for increasing temperature, and the single crystal hardness on its hardest orientation is only about half of the polycrystalline material depending on the test temperature. Because of its polycrystalline character, some cobalt bonded cemented tungsten carbides can be harder than single crystal WC over some intermediate temperature range.     

Source tracking of microbial intrusion in water systems using artificial neural networks

A "what-if" scenario where biological agents are accidentally or deliberately introduced into a water system was generated, and artificial neural network (ANN) models were applied to identify the pathogenic release location to isolate the contaminated area and minimize its hazards. The spatiotemporal distribution of Escherichia coli 15597 along the water system was employed to locate pollutants by inversely interpreting transport patterns of E. coli using ANNs. Results showed that dispersion patterns of E. coli were positively correlated to pH, turbidity, and conductivity (R2=0.90-0.96), and the ANN models successfully identified the source location of E. coli introduced into a given system with 75% accuracy based on the pre-programmed relationships between E. coli transport patterns and release locations. The findings in this study will enable us to assess the vulnerability of essential water systems, establish the early warning system and protect humans and the environment.     

Photomultiplication-Type Organic Photodetectors with High EQE-Bandwidth Product by Introducing a Perovskite Quantum Dot Interlayer

A photomultiplication (PM)-type organic photodetector (OPD) that exploits the ionic motion in CsPbI₃ perovskite quantum dots (QDs) is demonstrated. The device uses a QD monolayer as a PM-inducing interlayer and a donor–acceptor bulk heterojunction (BHJ) layer as a photoactive layer. When the device is illuminated, negative ions in the CsPbI₃ QD migrate and accumulate near the interface between the QDs and the electrode; these processes induce hole injection from the electrode and yield the PM phenomenon with an external quantum efficiency (EQE) >2000% at a 3 V applied bias. It is confirmed that the ionic motion of the CsPbI₃ QDs can induce a shift in the work function of the QD/electrode interface and that the dynamics of ionic motion determines the response speed of the device. The PM OPD showed a large EQE-bandwidth product >10⁶ Hz with a −3 dB frequency of 125 kHz at 3 V, which is one of the highest response speeds reported for a PM OPD. The PM-inducing strategy that exploits ionic motion of the interlayer is a potential approach to achieving high-efficiency PM OPDs.     

Sequence classification via large margin hidden Markov models

We address the sequence classification problem using a probabilistic model based on hidden Markov models (HMMs). In contrast to commonly-used likelihood-based learning methods such as the joint/conditional maximum likelihood estimator, we introduce a discriminative learning algorithm that focuses on class margin maximization. Our approach has two main advantages: (i) As an extension of support vector machines (SVMs) to sequential, non-Euclidean data, the approach inherits benefits of margin-based classifiers, such as the provable generalization error bounds. (ii) Unlike many algorithms based on non-parametric estimation of similarity measures that enforce weak constraints on the data domain, our approach utilizes the HMM’s latent Markov structure to regularize the model in the high-dimensional sequence space. We demonstrate significant improvements in classification performance of the proposed method in an extensive set of evaluations on time-series sequence data that frequently appear in data mining and computer vision domains.     

Explainable AI for domain experts: a post Hoc analysis of deep learning for defect classification of TFT–LCD panels

Journal of Intelligent Manufacturing - The deep learning (DL) model has performed successfully in various fields, including manufacturing. DL models for defect image data analysis in the...     

A Novel GPU Power Model for Accurate Smartphone Power Breakdown

As GPU power consumption in smartphones increases with more advanced graphic performance, it becomes essential to estimate GPU power consumption accurately. The conventional GPU power model assumes, simply, that a GPU consumes constant power when turned on; however, this is no longer true for recent smartphone GPUs. In this paper, we propose an accurate GPU power model for smartphones, considering newly adopted dynamic voltage and frequency scaling. For the proposed GPU power model, our evaluation results show that the error rate for system power estimation is as low as 2.9%, on average, and 4.6% in the worst case.