Journal of ICT Research and Applications
Not a member yet
359 research outputs found
Sort by
A CNN-ELM Classification Model for Automated Tomato Maturity Grading
Get PDFTomatoes are popular around the world due to their high nutritional value. Tomatoes are also one of the world’s most widely cultivated and profitable crops. The distribution and marketing of tomatoes depend highly on their quality. Estimating tomato ripeness is an essential step in determining shelf life and quality. With the abundant supply of tomatoes on the market, it is exceedingly difficult to estimate tomato ripeness using human graders. To address this issue and improve tomato quality inspection and sorting, automated tomato maturity classification models based on different features have been developed. However, current methods heavily rely on human-engineered or handcrafted features. Convolutional neural networks have emerged as the preferred technique for general object recognition problems because they can automatically detect and extract valuable features by directly working on input images. This paper proposes a CNN-ELM classification model for automated tomato maturity grading that combines CNNs’ automated feature learning capabilities with the efficiency of extreme learning machines to perform fast and accurate classification even with limited training data. The results showed that the proposed CNN-ELM model had a classification accuracy of 96.67% and an F1-score of 96.67% in identifying six maturity stages from the test data
Context-Aware Sentiment Analysis using Tweet Expansion Method
Get PDFThe large source of information space produced by the plethora of social media platforms in general and microblogging in particular has spawned a slew of new applications and prompted the rise and expansion of sentiment analysis research. We propose a sentiment analysis technique that identifies the main parts to describe tweet intent and also enriches them with relevant words, phrases, or even inferred variables. We followed a state-of-the-art hybrid deep learning model to combine Convolutional Neural Network (CNN) and the Long Short-Term Memory network (LSTM) to classify tweet data based on their polarity. To preserve the latent relationships between tweet terms and their expanded representation, sentence encoding and contextualized word embeddings are utilized. To investigate the performance of tweet embeddings on the sentiment analysis task, we tested several context-free models (Word2Vec, Sentence2Vec, Glove, and FastText), a dynamic embedding model (BERT), deep contextualized word representations (ELMo), and an entity-based model (Wikipedia). The proposed method and results prove that text enrichment improves the accuracy of sentiment polarity classification with a notable percentage
Translating SIBI (Sign System for Indonesian Gesture) Gesture-to-Text in Real-Time using a Mobile Device
Get PDFThe SIBI gesture translation framework by Rakun was built using a series of machine learning technologies: MobileNetV2 for feature extraction, Conditional Random Field for finding the epenthesis movement frame, and Long Short-Term Memory for word classification. This high computational translation system was previously implemented on a personal computer system, which lacks portability and accessibility. This study implemented the system on a smartphone using an on-device inference method: the translation process is embedded into the smartphone to provide lower latency and zero data usage. The system was then improved using a parallel multi-inference method, which reduced the average translation time by 25%. The final mobile SIBI gesture-to-text translation system achieved a word accuracy of 90.560%, a sentence accuracy of 64%, and an average translation time of 20 seconds
A New Term Frequency with Gaussian Technique for Text Classification and Sentiment Analysis
Get PDFThis paper proposes a new term frequency with a Gaussian technique (TF-G) to classify the risk of suicide from Thai clinical notes and to perform sentiment analysis based on Thai customer reviews and English tweets of travelers that use US airline services. This research compared TF-G with term weighting techniques based on Thai text classification methods from previous researches, including the bag-of-words (BoW), term frequency (TF), term frequency-inverse document frequency (TF-IDF), and term frequency-inverse corpus document frequency (TF-ICF) techniques. Suicide risk classification and sentiment analysis were performed with the decision tree (DT), naïve Bayes (NB), support vector machine (SVM), random forest (RF), and multilayer perceptron (MLP) techniques. The experimental results showed that TF-G is appropriate for feature extraction to classify the risk of suicide and to analyze the sentiments of customer reviews and tweets of travelers. The TF-G technique was more accurate than BoW, TF, TF-IDF and TF-ICF for term weighting in Thai suicide risk classification, for term weighting in sentiment analysis of Thai customer reviews for Burger King, Pizza Hut, and Sizzler restaurants, and for the sentiment analysis of English tweets of travelers using US airline services
Development of Hybrid Automatic Segmentation Technique of a Single Leaf from Overlapping Leaves Image
Get PDFThe segmentation of a single leaf from an image with overlapping leaves is an important step towards the realization of effective precision agricultural systems. A popular approach used for this segmentation task is the hybridization of the Chan-Vese model and the Sobel operator CV-SO. This hybridized approach is popular because of its simplicity and effectiveness in segmenting a single leaf of interest from a complex background of overlapping leaves. However, the manual threshold and parameter tuning procedure of the CV-SO algorithm often degrades its detection performance. In this paper, we address this problem by introducing a dynamic iterative model to determine the optimal parameters for the CV-SO algorithm, which we dubbed the Dynamic CV-SO (DCV-SO) algorithm. This is a new hybrid automatic segmentation technique that attempts to improve the detection performance of the original hybrid CV-SO algorithm by reducing its mean error rate. The results obtained via simulation indicate that the proposed method yielded a 1.23% reduction in the mean error rate against the original CV-SO method
Convolution and Recurrent Hybrid Neural Network for Hevea Yield Prediction
Get PDFDeep learning techniques have been used effectively for rubber crop yield prediction. A hybrid of Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) is the best technique for crop yield prediction because it can effectively handle uncertainty of features. Hence, in this paper, a hybrid CNN-RNN method is proposed to forecast Hevea yields based on environmental data in Kerala state, India. The proposed hybrid CNN-RNN method reduces the internal covariate shift of CNN by batch normalization and solves the gradient vanishing or exploding problem of RNN using LSTM with a cell activation mechanism. The proposed method has three essential characteristics: (i) it captures the time dependency of environmental factors and improves the inherent computational time; (ii) it is capable of generalizing the yield prediction under uncertain conditions without loss of prediction accuracy; (iii) combined with the back propagation and feed forward method it can reveal the extent to which samples of weather conditions and soil data conditions are suitable to provide a clear boundary between rubber yield variations
