Dr. Rahat Hossain Faisal - Academia.edu (original) (raw)

Papers by Dr. Rahat Hossain Faisal

Chapman and Hall/CRC eBooks, Jan 20, 2023

Garments and fashion industries play a vital role in our economy. The automatic classification an... more Garments and fashion industries play a vital role in our economy. The automatic classification and recognition of garments design class may help in development of fashion industry. For this purpose, different feature descriptors have been proposed to extract discriminative information from the garments texture images. In this paper we proposed a new descriptor namely Extended Noise Adaptive Binary Pattern (ENABP). To evaluate this descriptor, we use two different publicly available datasets (Fashion and Clothing attribute dataset). The experimental result shows that ENABP produces better accuracy than NABP and other existing descriptor.

International journal of information and electronics engineering, Dec 1, 2018

Different types of codes which may increase the liability of bugs or defects in future to a syste... more Different types of codes which may increase the liability of bugs or defects in future to a system known as Code smell. These type of smell can be eliminated without changing the external outcome and modifying the internal structure of the system. There are existing several well known code smell detection tools which automatically identify the code smells. The research used PMD automatic code smell detector to find four code smells on various open source java projects. It uses 43 open source java projects and identify the selective smells to these projects. The experiment shows duplicate codes, unused imports, unused local variables and unused private methods are not present for 79%, 34.9%, 51.1%, and 86.04% projects respectively. In the paper, it also shows that the probability of occurring duplicate codes, unused imports, unused local variables and unused private methods are respectively 4.5%, 71%, 20.4% and 4.1% which indicates the four selective code smells are declining in real life projects day by day.

Bangla is one of the mostly spoken languages all over the world, nonetheless, efforts on Bangla h... more Bangla is one of the mostly spoken languages all over the world, nonetheless, efforts on Bangla handwritten character recognition is not adequate. The use of Deep Convolutional Neural Network (DCNN) based classifiers has become a triumph over the state of art machine learning techniques. Using DCNN model to classify Bangla isolated basic characters can provide better output because of its ability to detect many hidden features from an image. In this paper, a new DCNN model, namely, BBCNet-15 is proposed for Bangla handwritten basic character recognition. the proposed model consists of 6 convolution layers, 6 max pooling layers, 2 fully connected layers followed by the softmax output unit. To avoid overfitting dropout regularization technique is used. The implementation of the proposed DCNN model is evaluated on a benchmark dataset, CMATERdb 3.1.2 which contains 50 character classes including 39 consonants and 11 vowels. The experimental evaluation of BBCNet-15 on the dataset provides a recognition accuracy of 96.40% which outperforms some prominent techniques in existence.

Applied Intelligence, Aug 16, 2021

From a real-world perspective, missing information is an ordinary scenario in data stream. Genera... more From a real-world perspective, missing information is an ordinary scenario in data stream. Generally, missing data generate diverse problems in recognizing the pattern of data (i.e., clustering and classification). Particularly, missing data in data stream is a challenging topic. With imbalanced data, the problem of missing data greatly affects pattern recognition. As a solution to all these issues, this study puts forward an adaptive technique with fuzzy-based information decomposition method, which simultaneously solves the problem of incomplete data and overcomes the imbalanced data stream in a dataset. The main purpose of the proposed fuzzy adaptive imputation approach (FAIA) is to represent the effect of missing values in imbalance data stream and handle the missing data problem in imbalance data stream. FAIA is a single pass method. It considers adaptive selection of intervals based on all observed instances by using the interrelationship of attributes to identify correct interval for computing missing instances. Here, the interrelationship of two attributes means one attribute’s value of an instance depends on another attribute’s value of the same instance. In FAIA, after measuring all interval distances from a certain missing value, the least distance is selected for this missing value. Synthetic data of minority class are generated using the same process of missing value imputation for balancing data that is called oversampling. Instances of the datasets are divided into the chunks in data stream to balance data without any ensemble of previous chunks because missing values may misguide the future chunks. To demonstrate the performance of FAIA, the experiment is divided into three parts: missing data imputation, imbalanced information for offline data for data stream, and imbalanced information with missing value for offline data. Eleven numerical datasets with different dimensions from various repositories are considered for the computing performance of missing data imputation and imbalanced data without data stream. Four different datasets are also used to measure the performance of imbalanced data stream. In maximum measuring cases, the proposed method outperforms.

Optical Character Recognition (OCR) especially for handwritten characters is an important task fo... more Optical Character Recognition (OCR) especially for handwritten characters is an important task for its numerous applications in daily life including data digitizing, robotics vision, helping visually disabled people and many more. However, Bangla Handwritten Character Recognition (HCR) is rarely explored despite Bangla being one of the mostly spoken languages over the world. For classifying Bangla basic characters, compound characters and digits various feature descriptors and classification algorithms can be used. This paper provides a comparative study of different Local Binary Pattern (LBP) based feature descriptors on Bangla basic characters, compound characters and digits. For classification, Support Vector Machine (SVM) with linear kernel is used. The rigorous experiments on CMATERdb 3.1.2, CMATERdb 3.1.3.1 and CMATERdb 3.1.1 datasets for Bangla basic characters, Bangla compound characters and Bangla digits respectively have showed reasonable accuracies of different LBP based feature descriptors.

Lecture notes in networks and systems, 2022

Deep Convolutional Neural Network has recently gained popularity because of its improved performa... more Deep Convolutional Neural Network has recently gained popularity because of its improved performance over the typical machine learning algorithms. However, it has been very rarely used on recognition of Bangla handwritten digit. This paper proposes a Deep Convolutional Neural Network (DCNN) based Bangla handwritten digits recognition scheme. The proposed method applies a seven layered D-CNN containing three convolution layers, three average pool layers and one fully connected layer for recognizing Bangla handwritten digits. Rigorous experimentation on a relatively large Bangla digit dataset namely, CMATERdb 3.1.1 provides considerable recognition accuracies.

2019 1st International Conference on Advances in Science, Engineering and Robotics Technology (ICASERT)

Hazy image enhancement is one of the challenging fields in the area of image processing because t... more Hazy image enhancement is one of the challenging fields in the area of image processing because the presence of haze reduces the visibility of outdoor images. In recent years, various algorithms have been introduced to remove haze. Most of the algorithms remove haze but hardly enhance the image with better contrast. In this paper, we have proposed an effective method for removing haze from the image. The main benefit of our method is that it dehazes image as well as preserves ramp edges providing better contrast of the image. The qualitative and quantitative comparison among few states of the art methods have been performed which show that our method gives better output images compared to those methods.

2019 International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering (IC4ME2)

In Computer Vision, automatic processing system gaining its popularity for its powerful classific... more In Computer Vision, automatic processing system gaining its popularity for its powerful classification and detection ability. Indigenous fish is an important element in natural food system which established the main diet in rural households. Hence, the classification of indigenous fish plays a vital role in authentication, preservation, and production. In this paper, we introduce a Hybrid Local Binary Pattern (HLBP), an adaptive threshold based hybrid feature descriptor which extracts sign and magnitude from an image. Afterward, we use different kernels of SVM for classification.We have also created a new indigenous fish dataset namely BDIndigenousFish2019 which contains images of eight different Bangladeshi fish species. The experimental result on BDIndigenousFish2019. The proposed HLBP is implemented for the classification of some indigenous fish species of Bangladesh with different kernels of SVM classifier. This paper focuses on the classification of some indigenous fishes of Bangladesh by means of SVM classifier with different kernels. We have conducted the experiment on our own indigenous fish dataset and comparative analysis HLBP with some well-known feature descriptors such as LBP, LGP, NABP, CENTRIST, DTCTH and LAID. Therefore, we evaluate the experimental results and our proposed model gain higher accuracy of 90% than other methods.

2019 International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering (IC4ME2), 2019

Proceedings of the Third International Conference on Trends in Computational and Cognitive Engineering, 2022

2019 Joint 8th International Conference on Informatics, Electronics & Vision (ICIEV) and 2019 3rd International Conference on Imaging, Vision & Pattern Recognition (icIVPR), 2019

2019 International Conference on Electrical, Computer and Communication Engineering (ECCE), 2019

For image classification, Local Gradient Pattern (LGP) is an adaptive threshold-based feature des... more For image classification, Local Gradient Pattern (LGP) is an adaptive threshold-based feature descriptor which extracts the changes of intensities locally or globally of an image. This threshold is calculated by using Arithmetic Mean (AM) of gradient values of neighboring pixels. Due to using AM, the threshold value often unable to reduce outlier's effect. Hence some of the elements of an image are not identified properly. As a result, the discrimination capacity of LGP comparatively lower than other descriptors for several applications. Above this issue, we introduce a new gradient-based feature descriptor named as modified Local Gradient Pattern (hLGP) to overcome this problem of LGP. This paper shows the effective performance of hLGP on several applications like scene, flower, aerial, event, object image classifications which belong to some popular datasets and also show the experimental results which exhibited that hLG P performs comparatively better than LGP in those datasets.

2019 International Conference on Electrical, Computer and Communication Engineering (ECCE), 2019

Early recognition and diagnosis of fruit Plants diseases are more important for increasing the de... more Early recognition and diagnosis of fruit Plants diseases are more important for increasing the development of healthy fruits industry. General observation system by farmers perhaps time-consuming, expensive and sometimes inaccurate. For this purpose, we introduce an appropriate deep convolution neural network(D-CNN) based approach for identifying guava leaf diseases automatically. The proposed model applies to classify major diseases of guava leaf such as Algal Leaf spot, Whitefly, and Rust. As per our insight, this is the first time to use D-CNN based model to recognize guava leaf diseases. Besides, we create our own dataset namely BU_Guava_Leaf (BUGL2018) with four different categories. On this dataset, we train our proposed disease identification approach and evaluate the experimental result which shows the average accuracy of 98.74% on test-set.

2018 21st International Conference of Computer and Information Technology (ICCIT), 2018

Irrigation is the process of applying appropriate amount of water to crop fields at needed interi... more Irrigation is the process of applying appropriate amount of water to crop fields at needed interim. Irrigation is an exigent part of cultivation of rice. Although, the overall yield of rice paddy predominantly depends on proper irrigation, irrigation process in developing countries like Bangladesh is still backdated. This study proposes an Arduino/GSM based remotely controlled power efficient smart irrigation system for crops that need to be immersed in water during its growing period. It will ensure the proper irrigation of a field by monitoring water level of the paddy field, providing feedback to farmers and giving farmers option to control the water motor via SMS. This study is expected to improve the overall production of AMAN and BORO rice of Barishal region by automating the traditional irrigation system. Also it will provide a more sophisticated irrigation system for similar types of crops.

2019 Joint 8th International Conference on Informatics, Electronics & Vision (ICIEV) and 2019 3rd International Conference on Imaging, Vision & Pattern Recognition (icIVPR), 2019

Most of the people in the world rely on traditional medicine which is made from medicinal plants.... more Most of the people in the world rely on traditional medicine which is made from medicinal plants. However, very few works concentrate on automatic classification. Therefore, the automatic classification of medicinal plants demands more investigation which is an important issue for conservation, authentication, and production of medicines. In this paper, for automatically classifying medicinal plants, we present a Multi-channel Modified Local Gradient Pattern (MCMLGP), a new texture-based feature descriptor that uses different channels of color images for extracting more significant features to improve the performance of classification. We have trained our proposed approach using SVM classifier with various kernels such as linear, polynomial and HI. In addition, we have used different feature descriptors for comparative experimental analysis with MCMLGP by conducting the rigorous experiment on our own medicinal plants dataset. The proposed approach gain higher accuracy (96.11%) than other techniques, and significantly valuable for exploration and evolution of medicinal plants classification.

Applied Intelligence, 2021

2019 International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering (IC4ME2), 2019

Aerial image classification has become one of the most important topics to the computer vision re... more Aerial image classification has become one of the most important topics to the computer vision researchers because of its numerous real world application. A great number of census transform based descriptors have been introduced in recent years to classify the aerial images. But the major drawback of these census transform based techniques is, most of these techniques works only with the center pixel information of an image with respect to their neighboring pixels. Hence, no information about the relationship among the neighboring pixels is obtained. To mitigate this problem, we introduce an Augmented Census Transform Histogram (ACENTRIST) for aerial image classification which encodes both the center pixel information and neighboring pixel information. The proposed technique augments two local binary pattern based descriptor which encodes the center pixel information with respect to the neighboring pixels and information of the angular difference of the neighboring pixels. We have conducted thorough experiments in two of the well-known aerial image dataset, UC Merced Land Use (Land Use 21) and In-House (Banja Luka), and the experimental result shows that the proposed methodology gains considerable higher accuracy over the state of the art methods.