RESUMO
Automated diagnosis of cancer from skin lesion data has been the focus of numerous research. Despite that it can be challenging to interpret these images because of features like colour illumination changes, variation in the sizes and forms of the lesions. To tackle these problems, the proposed model develops an ensemble of deep learning techniques for skin cancer diagnosis. Initially, skin imaging data are collected and preprocessed using resizing and anisotropic diffusion to enhance the quality of the image. Preprocessed images are fed into the Fuzzy-C-Means clustering technique to segment the region of diseases. Stacking-based ensemble deep learning approach is used for classification and the LSTM acts as a meta-classifier. Deep Neural Network (DNN) and Convolutional Neural Network (CNN) are used as input for LSTM. This segmented images are utilized to be input into the CNN, and the local binary pattern (LBP) technique is employed to extract DNN features from the segments of the image. The output from these two classifiers will be fed into the LSTM Meta classifier. This LSTM classifies the input data and predicts the skin cancer disease. The proposed approach had a greater accuracy of 97%. Hence, the developed model accurately predicts skin cancer disease.
RESUMO
The rapid advancement in AI requires efficient accelerators for training on edge devices, which often face challenges related to the high hardware costs of floating-point arithmetic operations. To tackle these problems, efficient floating-point formats inspired by block floating-point (BFP), such as Microsoft Floating Point (MSFP) and FlexBlock (FB), are emerging. However, they have limited dynamic range and precision for the smaller magnitude values within a block due to the shared exponent. This limits the BFP's ability to train deep neural networks (DNNs) with diverse datasets. This paper introduces the hybrid precision (HPFP) selection algorithms, designed to systematically reduce precision and implement hybrid precision strategies, thereby balancing layer-wise arithmetic operations and data path precision to address the shortcomings of traditional floating-point formats. Reducing the data bit width with HPFP allows more read/write operations from memory per cycle, thereby decreasing off-chip data access and the size of on-chip memories. Unlike traditional reduced precision formats that use BFP for calculating partial sums and accumulating those partial sums in 32-bit Floating Point (FP32), HPFP leads to significant hardware savings by performing all multiply and accumulate operations in reduced floating-point format. For evaluation, two training accelerators for the YOLOv2-Tiny model were developed, employing distinct mixed precision strategies, and their performance was benchmarked against an accelerator utilizing a conventional brain floating point of 16 bits (Bfloat16). The HPFP selection, employing 10 bits for the data path of all layers and for the arithmetic of layers requiring low precision, along with 12 bits for layers requiring higher precision, results in a 49.4% reduction in energy consumption and a 37.5% decrease in memory access. This is achieved with only a marginal mean Average Precision (mAP) degradation of 0.8% when compared to an accelerator based on Bfloat16. This comparison demonstrates that the proposed accelerator based on HPFP can be an efficient approach to designing compact and low-power accelerators without sacrificing accuracy.
RESUMO
Energy efficiency and security issues are the main concerns in wireless sensor networks (WSNs) because of limited energy resources and the broadcast nature of wireless communication. Therefore, how to improve the energy efficiency of WSNs while enhancing security performance has attracted widespread attention. In order to solve this problem, this paper proposes a new deep reinforcement learning (DRL)-based strategy, i.e., DeepNR strategy, to enhance the energy efficiency and security performance of WSN. Specifically, the proposed DeepNR strategy approximates the Q-value by designing a deep neural network (DNN) to adaptively learn the state information. It also designs DRL-based multi-level decision-making to learn and optimize the data transmission paths in real time, which eventually achieves accurate prediction and decision-making of the network. To further enhance security performance, the DeepNR strategy includes a defense mechanism that responds to detected attacks in real time to ensure the normal operation of the network. In addition, DeepNR adaptively adjusts its strategy to cope with changing network environments and attack patterns through deep learning models. Experimental results show that the proposed DeepNR outperforms the conventional methods, demonstrating a remarkable 30% improvement in network lifespan, a 25% increase in network data throughput, and a 20% enhancement in security measures.
RESUMO
Finding the most interesting areas of an image is the aim of saliency detection. Conventional methods based on low-level features rely on biological cues like texture and color. These methods, however, have trouble with processing complicated or low-contrast images. In this paper, we introduce a deep neural network-based saliency detection method. First, using semantic segmentation, we construct a pixel-level model that gives each pixel a saliency value depending on its semantic category. Next, we create a region feature model by combining both hand-crafted and deep features, which extracts and fuses the local and global information of each superpixel region. Third, we combine the results from the previous two steps, along with the over-segmented superpixel images and the original images, to construct a multi-level feature model. We feed the model into a deep convolutional network, which generates the final saliency map by learning to integrate the macro and micro information based on the pixels and superpixels. We assess our method on five benchmark datasets and contrast it against 14 state-of-the-art saliency detection algorithms. According to the experimental results, our method performs better than the other methods in terms of F-measure, precision, recall, and runtime. Additionally, we analyze the limitations of our method and propose potential future developments.
RESUMO
Improving system security can be achieved through people identification. Among various methods, electroencephalography-based (EEG-based) identification is a dependable way to prevent identity theft and impersonation. Due to the distractions present in the identification environment, such as lack of focus, mental engagement, small body movements, blinking, and other noises, it is essential to analyze data that reflects these conditions. The present research aims to advance practical EEG-based identification by studying data with mental preoccupation and developing a suitable algorithm. In this article, data from a study conducted on a group of 109 individuals has been analyzed. The data is categorized into two groups: focused data and waiting data. The article describes preprocessing the data and extracting three types of features, including Statistical, Frequency, and Wavelet. Then, a deep neural network (DNN) is used to classify the data. The DNN utilizes a multilayer, fully-connected neural network, with the number of layers and neurons varying based on the data type. Optimization and regularization methods are employed to improve the accuracy of the results. The DNN achieved an average accuracy of 99.19% for frequency features over all subjects in the focused data category, while the waiting data category showed an accuracy of 97.81%.
RESUMO
Dental age estimation is extensively employed in forensic medicine practice. However, the accuracy of conventional methods fails to satisfy the need for precision, particularly when estimating the age of adults. Herein, we propose an approach for age estimation utilizing orthopantomograms (OPGs). We propose a new dental dataset comprising OPGs of 27,957 individuals (16,383 females and 11,574 males), covering an age range from newborn to 93 years. The age annotations were meticulously verified using ID card details. Considering the distinct nature of dental data, we analyzed various neural network components to accurately estimate age, such as optimal network depth, convolution kernel size, multi-branch architecture, and early layer feature reuse. Building upon the exploration of distinctive characteristics, we further employed the widely recognized method to identify models for dental age prediction. Consequently, we discovered two sets of models: one exhibiting superior performance, and the other being lightweight. The proposed approaches, namely AGENet and AGE-SPOS, demonstrated remarkable superiority and effectiveness in our experimental results. The proposed models, AGENet and AGE-SPOS, showed exceptional effectiveness in our experiments. AGENet outperformed other CNN models significantly by achieving outstanding results. Compared to Inception-v4, with the mean absolute error (MAE) of 1.70 and 20.46 B FLOPs, our AGENet reduced the FLOPs by 2.7×. The lightweight model, AGE-SPOS, achieved an MAE of 1.80 years with only 0.95 B FLOPs, surpassing MobileNetV2 by 0.18 years while utilizing fewer computational operations. In summary, we employed an effective DNN searching method for forensic age estimation, and our methodology and findings hold significant implications for age estimation with oral imaging.
RESUMO
This paper introduces a mobile cloud-based predictive model for assisting Parkinson's disease (PD) patients. PD, a chronic neurodegenerative disorder, impairs motor functions and daily tasks due to the degeneration of dopamine-producing neurons in the brain. The model utilizes smartphones to aid patients in collecting voice samples, which are then sent to a cloud service for storage and processing. A hybrid deep learning model, trained using the UCI Parkinson's Telemonitoring Voice dataset, analyzes this data to estimate the severity of PD symptoms. The model's performance is noteworthy, with accuracy, sensitivity, and specificity metrics of 96.2 %, 94.15 %, and 96.15 %, respectively. Additionally, it boasts a rapid response time of just 13 s. Results are delivered to users via smartphone alert notifications, coupled with a knowledge base feature that educates them about PD. This system provides reliable home-based assessment and monitoring of PD and enables prompt medical intervention, significantly enhancing the quality of life for patients with Parkinson's disease.
RESUMO
In a recent paper by Sajindra et al. [1], the soil nutrient levels, specifically nitrogen, phosphorus, and potassium, in organic cabbage cultivation were predicted using a deep learning model. This model was designed with a total of four hidden layers, excluding the input and output layers, with each hidden layer meticulously crafted to contain ten nodes. The selection of the tangent sigmoid transfer function as the optimal activation function for the dataset was based on considerations such as the coefficient of correlation, mean squared error, and the accuracy of the predicted results. Throughout this study, the objective is to justify the tangent sigmoid transfer function and provide mathematical justification for the obtained results.â¢This paper presents the comprehensive methodology for the development of deep neural network for predict the soil nutrient levels.â¢Tangent Sigmoid transfer function usage is justified in predictions.â¢Methodology can be adapted to any similar real-world scenarios.
RESUMO
Rice is a staple crop in Asia, with more than 400 million tons consumed annually worldwide. The protein content of rice is a major determinant of its unique structural, physical, and nutritional properties. Chemical analysis, a traditional method for measuring rice's protein content, demands considerable manpower, time, and costs, including preprocessing such as removing the rice husk. Therefore, of the technology is needed to rapidly and nondestructively measure the protein content of paddy rice during harvest and storage stages. In this study, the nondestructive technique for predicting the protein content of rice with husks (paddy rice) was developed using near-infrared spectroscopy and deep learning techniques. The protein content prediction model based on partial least square regression, support vector regression, and deep neural network (DNN) were developed using the near-infrared spectrum in the range of 950 to 2200 nm. 1800 spectra of the paddy rice and 1200 spectra from the brown rice were obtained, and these were used for model development and performance evaluation of the developed model. Various spectral preprocessing techniques was applied. The DNN model showed the best results among three types of rice protein content prediction models. The optimal DNN model for paddy rice was the model with first-order derivative preprocessing and the accuracy was a coefficient of determination for prediction, Rp 2 = 0.972 and root mean squared error for prediction, RMSEP = 0.048%. The optimal DNN model for brown rice was the model applied first-order derivative preprocessing with Rp 2 = 0.987 and RMSEP = 0.033%. These results demonstrate the commercial feasibility of using near-infrared spectroscopy for the non-destructive prediction of protein content in both husked rice seeds and paddy rice.