Title: Improving the Performance of CNN by Using Dominant Patterns of CNN for Hand Detection
<br>Cover Date: 2023-06-01
<br>Cover Display Date: June 2023
<br>DOI: 10.37936/ecti-cit.2023172.251265
<br>Description: Many applications have used hand gestures for software interaction, image-and video-based action analysis, and behavioral monitoring. Hand detection is an essential step in the pipeline of these applications, and Convolutional Neural Networks (CNN) has provided superior solutions. However, CNN has similar features between hand and non-hand images, called non-dominant features. These features affect miss-classifications and long-time computation. Therefore, this paper focuses on the selection of dominant CNN features for hand detection, and it is our proposed method (DP-CNN) that selects the dominant feature patterns (DP) from the trained CNN features and classifies them using the Extreme Learning Machine (ELM) method. Evaluation results show the proposed method (DP-CNN-ELM), which can increase the accuracy and the F1-score of CNN. In addition, the proposed method can reduce the time computation of CNN in training and testing.
<br>Citations: 2
<br>Aggregation Type: Journal
<br>-------------------
<br><br><br>Title: Improving Performance of Sparse Autoencoder by Using DPHOG for Gender Classification
<br>Cover Date: 2021-01-01
<br>Cover Display Date: 2021
<br>DOI: 10.1109/ICSEC53205.2021.9684578
<br>Description: This paper proposes the improvement performance of Sparse Auto Encoder (SAE) by using Dominant Patterns of Histograms of Oriented Gradients (DPHOG). The SAE has a simple structure and fast computation. However, the SAE demonstrates less accuracy than the Convolution of Neuron Networks (CNN). This proposed method selects the dominant features of female and male face images, and then encoding and decoding these features within SAE. The researchers conducted DPHOG with SAE in two datasets and compared them with SAE, HOG with SAE, and CNN. Experimental results showed that DPHOG with SAE produced the highest performance in terms of accuracy, true positive rate, false positive rate and time computation in the two datasets. In addition, the DPHOG with SAE can extract learning features the form small dataset. In contrast, CNN requires a huge dataset for learning features.
<br>Citations: 0
<br>Aggregation Type: Conference Proceeding
<br>-------------------
<br><br><br>Title: The Combination of Different Cell Sizes of HOG with KELM for Vehicle Detection
<br>Cover Date: 2020-01-01
<br>Cover Display Date: 2020
<br>DOI: 10.1007/978-3-030-19861-9_18
<br>Description: HOG has been developed successfully in many intelligent vehicle detection systems. HOG still has interesting problems that consist of (i) redundant features and (ii) ambiguous features (similarities between vehicles and non-vehicles), which problems have an effect on time computation and misclassification. The vertical direction of HOG method (V-HOG) and adding the position of orientation bins and intensity features (πHOG) improve the problems of HOG; but they produce redundant and ambiguous features in various regions of vehicles. This paper proposes a new method for improving the performance of HOG that has flexibility in various regions of vehicles. The proposed method used combines different sized cells of HOG that is called CDC-HOG. The CDC-HOG were conducted on a GTI dataset, which consists of four regions (far, front, left, and right regions). The CDC-HOG is compared with HOG, V-HOG, πHOG, and PHOG; uses the kernel extreme learning machine (KELM), and supports vector machine (SVM) for evaluating features. The CDC-HOG with KELM produced the highest performance in terms of accuracy, true positive rate, and false positive rate for all regions.
<br>Citations: 1
<br>Aggregation Type: Book Series
<br>-------------------
<br><br><br>Title: A Novel Feature Selection in Vehicle Detection Through the Selection of Dominant Patterns of Histograms of Oriented Gradients (DPHOG)
<br>Cover Date: 2019-01-01
<br>Cover Display Date: 2019
<br>DOI: 10.1109/ACCESS.2019.2893320
<br>Description: This paper proposes a novel method that addresses the selection of the dominant patterns of the histograms of oriented gradients (DPHOGs) in vehicle detection. HOG features lead to an expensive classification with high misclassification rates since HOG generates a long vector containing both redundant and ambiguous features (similarities between the vehicle and non-vehicle images). Several modifications of HOG were proposed to resolve these issues such as the vertical histograms of oriented gradient and one that includes position and intensity with HOG; however, these methods still contain some ambiguous features. A feature selection method can exclude these ambiguous features, allowing for better classification rates and a reduction in classification times. The proposed method uses the ideal vectors of the vehicle and non-vehicles images for selecting features in dominant patterns. The segments indicating the differences between the vehicle and non-vehicle classes are the dominant patterns, in which the length of the feature vector is shortened. We performed DPHOG on three standard datasets, in which the kernel extreme learning machine, the support vector machine, K-nearest neighbor, random forest, and deep neural network were used as classifiers. We then compared the performance of the DPHOG with eight well-known feature selection methods and three existing feature extraction methods for vehicle detection. In evaluations with each comparative method concerning the accuracy, true positive, false positive, and F1-score, the DPHOG presented the highest performances with less running time in each dataset.
<br>Citations: 14
<br>Aggregation Type: Journal
<br>-------------------
<br><br><br>Title: Comparative study of computational time that HOG-based features used for vehicle detection
<br>Cover Date: 2018-01-01
<br>Cover Display Date: 2018
<br>DOI: 10.1007/978-3-319-60663-7_26
<br>Description: HOG produces a number of redundant and long features so that they affect to the detection rate and computational time. This paper studied the processes that HOG-based features were generated, selected, and used in vehicle detection and find one that takes the shortest time. There were five combinations of feature extractors and classifiers. Time spent in HV step, accuracy of detection and the false positive rate are considered together for making decision of which combination is the best. The experiments were conducted on GIT dataset. The experimental results showed that process which VHOG preceded ELM provided a little less accurate than HOG preceded SVM did. However, it did not only take shortest time in HV step but also provided the lowest false positive rate. Therefore, VHOG preceded ELM should be selected as a method for vehicle detection.
<br>Citations: 17
<br>Aggregation Type: Book Series
<br>-------------------
<br><br><br>Title: The automatic Thai basketry detection and recognition on the local wisdom video
<br>Cover Date: 2015-02-09
<br>Cover Display Date: 9 February 2015
<br>DOI: 10.1109/INFOS.2014.7036707
<br>Description: To retrieve the videos having content relating to Local Thai basketry, creating video's indexes is vital because indexes allow users to more quickly find video for specific individuals. In this work, it aims to propose a novel methodology of video indexing. The proposed methodology consists of two main processing stages. The first stage is to capture video frames having the Thai basketry, and then classify them into many individual groups, because local Thai basketry can be classified into many types depending on its usability and pattern. This stage is driven on using color and texture, which are processed through the Artificial Neural networks (ANNs). The second stage is to recognize the basketry shape by chain code and template matching analysis on the object's shape in order to use them as video indexing. Finally, the proposed methodology is experimented on 41 local wisdom videos having 76 shorts and 4,196 frames. After testing by recall, precision, and F-measure, they show the satisfactory results for recall, precision, and F-measure as 78.64%, 83.88%, and 81.18%, respectively.
<br>Citations: 0
<br>Aggregation Type: Conference Proceeding
<br>-------------------
<br><br><br>Title: Improving vehicle detection by adapting parameters of HOG and kernel functions of SVM
<br>Cover Date: 2014-01-01
<br>Cover Display Date: 2014
<br>DOI: 10.1109/ICSEC.2014.6978225
<br>Description: Currently, vehicle detection suffers from low performance in terms of accuracy and time costs in real-time application. Histograms Oriented of Gradients(HOG) and Support Vector Machine(SVM) are popular methods used to address these problems, however, while they can give high accuracy, detection is still too slow for real-time application. The V-HOG method has previously been proposed to reduce detection time in real-time application by adjusting HOG structures. Although V-HOG detection is faster than that of HOG, the accuracy is lower. Therefore, this paper proposed to improve accuracy and classification time by adjusting HOG parameters and SVM kernel functions. The experimental results showed that the proposed method results in 100% accuracy and supports real-time application.
<br>Citations: 8
<br>Aggregation Type: Conference Proceeding
<br>-------------------
<br><br><br>Title: Images Enhancement of G-band Chromosome Using histogram equalization, OTSU thresholding, morphological dilation and flood fill techniques
<br>Cover Date: 2012-12-01
<br>Cover Display Date: 2012
<br>DOI: N/A
<br>Description: Human chromosomes contain important information for cytogenetic analysis. Cytogenetic compares their patient's chromosome images against the prototype human chromosome band patterns. Chromosome images were acquired by microscopic imaging of metaphase or prophase cells on specimen slides. Digitized chromosome images usually are suffered from poor image quality, lack of contrast and hole in images. Therefore, the images must be enhanced. This paper presented an enhancement algorithm for chromosome images based on histogram equalization (HE) OTSU flood-full and dilations. Firstly, the histogram equalization is used to improve the contrast of the image. Then the OTSU threshold segment, flood-fill fill hole in image and dilation are experimented to validate the effect of this algorithm, the canny edge detection and oriented bounding box were used to detect the edge and segment chromosomes from background. A success rate of the proposed method can be improved to 97.07% of the enhancement in the chromosome images. © 2012 AICIT.
<br>Citations: 23
<br>Aggregation Type: Conference Proceeding
<br>-------------------
<br><br><br>Title: A hybrid method for enhancement of G-banding metaphase chromosome
<br>Cover Date: 2012-11-01
<br>Cover Display Date: November 2012
<br>DOI: 10.4156/jcit.vol7.issue21.48
<br>Description: Human chromosomes contain important information for cytogenetic analysis. Cytogenetic compares their patient's chromosome images against the prototype human chromosome band patterns. Chromosome images were acquired by microscopic imaging of metaphase or prophase cells on specimen slides. Digitized chromosome images usually are suffered from poor image quality, lack of contrast and hole in images. Therefore, the images must be enhanced. This paper presented an enhancement algorithm for chromosome images based on histogram equalization (HE) Otsu thresholding flood-full and dilations. Firstly, the histogram equalization is used to improve the contrast of the image. Then, the Otsu threshold segment, flood-fill fill hole in image and dilation are experimented to validate the effect of this algorithm, the canny edge detection and oriented bounding box were used to detect the edge and segment chromosomes from background. A success rate of the proposed method can improve the enhancement in the chromosome images is 97. 07% has been achieved.
<br>Citations: 2
<br>Aggregation Type: Journal
<br>-------------------
<br><br><br>