Search

Human–object interaction (HOI) detection is important for promoting the development of many fields such as human–computer interactions, service robotics, and video security surveillance. A high percentage of human–object pairs with invalid interactions are discovered in the object detection phase of conventional human–object interaction detection algorithms, resulting in inaccurate interaction detection. To recognize invalid human–object interaction pairs, this paper proposes a model structure, the interactivity recognition graph neural network (IR-GNN) model, which can directly infer the probability of human–object interactions from a graph model architecture. The model consists of three modules: The first one is the human posture feature module, which uses key points of the human body to construct relative spatial pose features and further facilitates the discrimination of human–object interactivity through human pose information. Second, a human–object interactivity graph module is proposed. The spatial relationship of human–object distance is used as the initialization weight of edges, and the graph is updated by combining the message passing of attention mechanism so that edges with interacting node pairs obtain higher weights. Thirdly, the classification module is proposed; by finally using a fully connected neural network, the interactivity of human–object pairs is binarily classified. These three modules work in collaboration to enable the effective inference of interactive possibilities. On the datasets HICO-DET and V-COCO, comparative and ablation experiments are carried out. It has been proved that our technology can improve the detection of human–object interactions. [ABSTRACT FROM AUTHOR]

Breast cancer detection using mammogram images at an early stage is an important step in disease diagnostics. We propose a new method for the classification of benign or malignant breast cancer from mammogram images. Hybrid thresholding and the machine learning method are used to derive the region of interest (ROI). The derived ROI is then separated into five different blocks. The wavelet transform is applied to suppress noise from each produced block based on BayesShrink soft thresholding by capturing high and low frequencies within different sub-bands. An improved fractal dimension (FD) approach, called multi-FD (M-FD), is proposed to extract multiple features from each denoised block. The number of features extracted is then reduced by a genetic algorithm. Five classifiers are trained and used with the artificial neural network (ANN) to classify the extracted features from each block. Lastly, the fusion process is performed on the results of five blocks to obtain the final decision. The proposed approach is tested and evaluated on four benchmark mammogram image datasets (MIAS, DDSM, INbreast, and BCDR). We present the results of single- and double-dataset evaluations. Only one dataset is used for training and testing in the single-dataset evaluation, whereas two datasets (one for training, and one for testing) are used in the double-dataset evaluation. The experiment results show that the proposed method yields better results on the INbreast dataset in the single-dataset evaluation, whilst better results are obtained on the remaining datasets in the double-dataset evaluation. The proposed approach outperforms other state-of-the-art models on the Mini-MIAS dataset. [ABSTRACT FROM AUTHOR]

In this study, we proposed a new approach in estimating a minimum value of machining performance. In this approach, artificial neural network (ANN) and genetic algorithm (GA) techniques were integrated in order to search for a set of optimal cutting condition points that leads to the minimum value of machining performance. Three machining cutting conditions for end milling operation that were considered in this study are speed (v), feed (f) and radial rake angle (γ). The considered machining performance is surface roughness (R a). The minimum R a value at the optimal v, f and γ points was expected from this approach. Using the proposed approach, named integrated ANN–GA, this study has proven that R a can be estimated to be 0.139 µm, at the optimal cutting conditions of f = 167.029 m/min, v = 0.025 mm/tooth and γ = 14.769°. Consequently, the ANN–GA integration system has reduced the R a value at about 26.8%, 25.7%, 26.1% and 49.8%, compared to the experimental, regression, ANN and response surface method results, respectively. Compared to the conventional GA result, it was also found that integrated ANN–GA reduced the mean R a value and the number of iterations in searching for the optimal result at about 0.61% and 23.9%, respectively. [ABSTRACT FROM PUBLISHER]

As surveyed, many efforts have been made to model the performances of electrical discharge machining (EDM) using artificial neural network (ANN). However, the selections of the network parameters were mostly prepared in a random manner, resulting to unnecessary trials. Thus, orthogonal array (Taguchi) is employed in the procedure of network function and network architecture assortment to avoid excessive random trial experimentations. This proposed orthogonal based ANN modelling is employed on WEDM of Ti–48Al intermetallic alloys. Meanwhile modified multi objective genetic algorithm (multiGA) is used as the optimization technique. Material removal rate (MRR), surface roughness (Ra), cutting speed (Vc) and width of kerf (Dk) are the machining performances considered in this study. Five machining parameters observed from the previous researches are chosen as significant factors to the machining performances in this study, which are pulse on time, pulse off time, peak current, feed rate and servo voltage. Experimental studies are carried out to verify the machining performances suggested by this approach. Feed forward back propagation neural network (FFNN) is found to be the best network type on the selected dataset. Two hidden layer 5–6–6–4 FFNN showed the most precise and generalized network architecture with very good prediction accuracy. The proposed approach, OrthoANN, reduced ANN experimentation time by a large scale and produced viable results for machining optimization when integrated with multiGA. [ABSTRACT FROM AUTHOR]

Artificial Neural Network which possessing self-learning ability has shown great promise in addressing problem of learning an artist style in generating facial caricatures. This paper presents Artificial Neural Network model to imitate a particular artist style and to predict a facial caricature configuration for a given original face image (photo). This paper also describes the data preparing process that proposes a modified procrustes superimposition method in deriving the datasets for the neural network model. The experiment is carried out to compare the modified procrustes superimposition method with the original one and to find the appropriate neural network structure that would yield the most accurate prediction results. Different datasets (N1, N2, and N3) derived from the same raw data but using the different method in preparing the dataset and different numbers of hidden nodes (6, 12, 18 and 24) are tested. The experimental result and its detail analysis are given and discussed. It proves that neural network has an ability to predict how the artist exaggerates the original facial feature point. Dataset N2 which use Modified Procrustes Superimposition method and the simple structure of a single hidden layer neural network, in which 6 is the number of hidden node, give the best accuracy of the prediction. [ABSTRACT FROM AUTHOR]

Abstract: Surface roughness is one of the most common performance measurements in machining process and an effective parameter in representing the quality of machined surface. The minimization of the machining performance measurement such as surface roughness (Ra ) must be formulated in the standard mathematical model. To predict the minimum Ra value, the process of modeling is taken in this study. The developed model deals with real experimental data of the Ra in the end milling machining process. Two modeling approaches, regression and Artificial Neural Network (ANN), are applied to predict the minimum Ra value. The results show that regression and ANN models have reduced the minimum Ra value of real experimental data by about 1.57% and 1.05%, respectively. [Copyright &y& Elsevier]

Abstract: In this study, Artificial Neural Network (ANN) and Simulated Annealing (SA) techniques were integrated labeled as integrated ANN-SA to estimate optimal process parameters in abrasive waterjet (AWJ) machining operation. The considered process parameters include traverse speed, waterjet pressure, standoff distance, abrasive grit size and abrasive flow rate. The quality of the cutting of machined-material is assessed by looking to the roughness average value (R a). The optimal values of the process parameters are targeted for giving a minimum value of R a. It was evidence that integrated ANN-SA is capable of giving much lower value of R a at the recommended optimal process parameters compared to the result of experimental and ANN single-based modeling. The number of iterations for the optimal solutions is also decreased compared to the result of SA single-based optimization. [Copyright &y& Elsevier]

Abstract: This paper presents the ANN model for predicting the surface roughness performance measure in the machining process by considering the Artificial Neural Network (ANN) as the essential technique for measuring surface roughness. A revision of several previous studies associated with the modelling issue is carried out to assess how capable ANN is as a technique to model the problem. Based on the studies conducted by previous researchers, the abilities and limitations of the ANN technique for predicting surface roughness are highlighted. Utilization of ANN-based modelling is also discussed to show the required basic elements for predicting surface roughness in the milling process. In order to investigate how capable the ANN technique is at estimating the prediction value for surface roughness, a real machining experiment is referred to in this study. In the experiment, 24 samples of data concerned with the milling operation are collected based on eight samples of data of a two-level DOE full factorial analysis, four samples of centre data, and 12 samples of axial data. All data samples are tested in real machining by using uncoated, TiAIN coated and coated cutting tools of titanium alloy (Ti-6A1-4V). The Matlab ANN toolbox is used for the modelling purpose with some justifications. Feedforward backpropagation is selected as the algorithm with traingdx, learngdx, MSE, logsig as the training, learning, performance and transfer functions, respectively. With three nodes in the input layer and one node in the output layer, eight networks are developed by using different numbers of nodes in the hidden layer which are 3–1–1, 3–3–1, 3–6–1, 3–7–1, 3–1–1–1, 3–3–3–1, 3–6–6–1 and 3–7–7–1 structures. It was found that the 3–1–1 network structure of the coated cutting tool gave the best ANN model in predicting the surface roughness value. This study concludes that the model for surface roughness in the milling process could be improved by modifying the number of layers and nodes in the hidden layers of the ANN network structure, particularly for predicting the value of the surface roughness performance measure. As a result of the prediction, the recommended combination of cutting conditions to obtain the best surface roughness value is a high speed with a low feed rate and radial rake angle. [Copyright &y& Elsevier]