Pub Date : 2017-09-01DOI: 10.1109/ICSIPA.2017.8120598
Yassine Lehiani, Madjid Maidi, M. Preda, F. Ghorbel
In this paper, we present a novel approach for real-time object identification on a mobile platform. First, our system detects keypoints within a scaled pyramid-based FAST detector and then descriptors of the object of interest are computed using an Analytical Fourier-Mellin transform. The Fourier-Mellin is used in similarity studies due to its invariance property and discrimination power. In this approach, we exploited information from the phase of Fourier Transform instead of magnitude applied on patches. The phase carries more information and handle, particularly, rotation and light changes. Finally, experiments are conducted to evaluate the system performances in terms of accuracy, robustness and computational efficiency as well.
{"title":"Image invariant description based on local Fourier-Mellin transform","authors":"Yassine Lehiani, Madjid Maidi, M. Preda, F. Ghorbel","doi":"10.1109/ICSIPA.2017.8120598","DOIUrl":"https://doi.org/10.1109/ICSIPA.2017.8120598","url":null,"abstract":"In this paper, we present a novel approach for real-time object identification on a mobile platform. First, our system detects keypoints within a scaled pyramid-based FAST detector and then descriptors of the object of interest are computed using an Analytical Fourier-Mellin transform. The Fourier-Mellin is used in similarity studies due to its invariance property and discrimination power. In this approach, we exploited information from the phase of Fourier Transform instead of magnitude applied on patches. The phase carries more information and handle, particularly, rotation and light changes. Finally, experiments are conducted to evaluate the system performances in terms of accuracy, robustness and computational efficiency as well.","PeriodicalId":268112,"journal":{"name":"2017 IEEE International Conference on Signal and Image Processing Applications (ICSIPA)","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125282026","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-09-01DOI: 10.1109/ICSIPA.2017.8120604
Arushi Mishra, A. Krishna, D. Madathil
An innovative and cost effective lower extremity blood circulation enhancer works on the basic principle of whole body vibration. A prototype is designed with the circuitry consisting of five motors connected in parallel. The prototype is designed in the form of a strapped stocking which is to be worn on the legs and the foot. The motors have been positioned in the prototype in such a way that the major muscles of the lower extremity will directly receive the vibration. The prototype is designed using cotton cloth due to the advantages the material offers. After successful testing of the prototype on ten subjects, signals were acquired from the posterior tibial artery of each of them. The output was obtained in the form of graphs using MATLAB. The prototype was designed successfully at a low price using readily available materials. The outcome showed that there was an increase in circulation of blood in the lower extremities. Further, we also observed that human muscles behave like a spring complex. They readily absorb vibration. This can also be seen from the results of the signal processing.
{"title":"An innovative and cost effective lower extremity blood circulation enhancer","authors":"Arushi Mishra, A. Krishna, D. Madathil","doi":"10.1109/ICSIPA.2017.8120604","DOIUrl":"https://doi.org/10.1109/ICSIPA.2017.8120604","url":null,"abstract":"An innovative and cost effective lower extremity blood circulation enhancer works on the basic principle of whole body vibration. A prototype is designed with the circuitry consisting of five motors connected in parallel. The prototype is designed in the form of a strapped stocking which is to be worn on the legs and the foot. The motors have been positioned in the prototype in such a way that the major muscles of the lower extremity will directly receive the vibration. The prototype is designed using cotton cloth due to the advantages the material offers. After successful testing of the prototype on ten subjects, signals were acquired from the posterior tibial artery of each of them. The output was obtained in the form of graphs using MATLAB. The prototype was designed successfully at a low price using readily available materials. The outcome showed that there was an increase in circulation of blood in the lower extremities. Further, we also observed that human muscles behave like a spring complex. They readily absorb vibration. This can also be seen from the results of the signal processing.","PeriodicalId":268112,"journal":{"name":"2017 IEEE International Conference on Signal and Image Processing Applications (ICSIPA)","volume":"50 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124013032","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-09-01DOI: 10.1109/ICSIPA.2017.8120622
S. Hassan, R. Ibrahim, N. Saad, V. Asirvadam, Kishore Bingi, Tran Duc Chung
The predictive proportional integral (PPI) controller being simple and non conservative like its Smith predictor and internal model control (IMC) counterparts, has the ability to be employed in environments characterised by uncertainties such as stochastic network delays. However, its performance degrades in the presence of measurement noise. This is due to the prediction involve in the controller which acts like the derivative term of PID controller that amplifies high frequency noise. This necessitated the need for additional filtering of the noisy signals. Therefore, this work proposes three several filter configurations for PPI controller in a wireless environment. Simulation results shows that placing the filter within the PPI controller gives smooth control signal that can yield better control performance compared to both feedback and cascade filter configurations.
{"title":"Signal noise filter structure selection for predictive PI controller in a wireless networked control setting","authors":"S. Hassan, R. Ibrahim, N. Saad, V. Asirvadam, Kishore Bingi, Tran Duc Chung","doi":"10.1109/ICSIPA.2017.8120622","DOIUrl":"https://doi.org/10.1109/ICSIPA.2017.8120622","url":null,"abstract":"The predictive proportional integral (PPI) controller being simple and non conservative like its Smith predictor and internal model control (IMC) counterparts, has the ability to be employed in environments characterised by uncertainties such as stochastic network delays. However, its performance degrades in the presence of measurement noise. This is due to the prediction involve in the controller which acts like the derivative term of PID controller that amplifies high frequency noise. This necessitated the need for additional filtering of the noisy signals. Therefore, this work proposes three several filter configurations for PPI controller in a wireless environment. Simulation results shows that placing the filter within the PPI controller gives smooth control signal that can yield better control performance compared to both feedback and cascade filter configurations.","PeriodicalId":268112,"journal":{"name":"2017 IEEE International Conference on Signal and Image Processing Applications (ICSIPA)","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128186418","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-09-01DOI: 10.1109/ICSIPA.2017.8120611
H. Tan, K. Lim, H. Harno
Stochastic Diagonal Approximate Greatest Descent (SDAGD) is proposed to manage the optimization in two stages, (a) apply a radial boundary to estimate step length when the weights are far from solution, (b) apply Newton method when the weights are within the solution level set. This is inspired by a multi-stage decision control system where different strategies is used at different conditions. In numerical optimization context, larger steps should be taken at the beginning of optimization and gradually reduced when it is near to the minimum point. Nevertheless, the intuition of determining the radial boundary when the optimized parameters are far from the solution is yet to be investigated for high dimensional data. Radial step length in SDAGD manipulates the relative step length for iteration construction. SDAGD is implemented in a two layer Multilayer Perceptron to evaluate the effects of R on artificial neural networks. It is concluded that the greater the value of R, the higher the learning rate of SDAGD algorithm when the value of R is constrained in between 100 to 10,000.
{"title":"Radial effect in stochastic diagonal approximate greatest descent","authors":"H. Tan, K. Lim, H. Harno","doi":"10.1109/ICSIPA.2017.8120611","DOIUrl":"https://doi.org/10.1109/ICSIPA.2017.8120611","url":null,"abstract":"Stochastic Diagonal Approximate Greatest Descent (SDAGD) is proposed to manage the optimization in two stages, (a) apply a radial boundary to estimate step length when the weights are far from solution, (b) apply Newton method when the weights are within the solution level set. This is inspired by a multi-stage decision control system where different strategies is used at different conditions. In numerical optimization context, larger steps should be taken at the beginning of optimization and gradually reduced when it is near to the minimum point. Nevertheless, the intuition of determining the radial boundary when the optimized parameters are far from the solution is yet to be investigated for high dimensional data. Radial step length in SDAGD manipulates the relative step length for iteration construction. SDAGD is implemented in a two layer Multilayer Perceptron to evaluate the effects of R on artificial neural networks. It is concluded that the greater the value of R, the higher the learning rate of SDAGD algorithm when the value of R is constrained in between 100 to 10,000.","PeriodicalId":268112,"journal":{"name":"2017 IEEE International Conference on Signal and Image Processing Applications (ICSIPA)","volume":"28 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125502535","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-09-01DOI: 10.1109/ICSIPA.2017.8120668
Nihal Murali, Kunal Gupta, S. Bhanot
Training of artificial neural networks (ANNs) using reinforcement learning (RL) techniques is being widely discussed in the robot learning literature. The high model complexity of ANNs along with the model-free nature of RL algorithms provides a desirable combination for many robotics applications. There is a huge need for algorithms that generalize using raw sensory inputs, such as vision, without any hand-engineered features or domain heuristics. In this paper, the standard control problem of line following robot was used as a test-bed, and an ANN controller for the robot was trained on images from a live video feed using Q-learning. A virtual agent was first trained in simulation environment and then deployed onto a robot's hardware. The robot successfully learns to traverse a wide range of curves and displays excellent generalization ability. Qualitative analysis of the evolution of policies, performance and weights of the network provide insights into the nature and convergence of the learning algorithm.
{"title":"Analysis of Q-learning on ANNs for robot control using live video feed","authors":"Nihal Murali, Kunal Gupta, S. Bhanot","doi":"10.1109/ICSIPA.2017.8120668","DOIUrl":"https://doi.org/10.1109/ICSIPA.2017.8120668","url":null,"abstract":"Training of artificial neural networks (ANNs) using reinforcement learning (RL) techniques is being widely discussed in the robot learning literature. The high model complexity of ANNs along with the model-free nature of RL algorithms provides a desirable combination for many robotics applications. There is a huge need for algorithms that generalize using raw sensory inputs, such as vision, without any hand-engineered features or domain heuristics. In this paper, the standard control problem of line following robot was used as a test-bed, and an ANN controller for the robot was trained on images from a live video feed using Q-learning. A virtual agent was first trained in simulation environment and then deployed onto a robot's hardware. The robot successfully learns to traverse a wide range of curves and displays excellent generalization ability. Qualitative analysis of the evolution of policies, performance and weights of the network provide insights into the nature and convergence of the learning algorithm.","PeriodicalId":268112,"journal":{"name":"2017 IEEE International Conference on Signal and Image Processing Applications (ICSIPA)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134476691","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-09-01DOI: 10.1109/ICSIPA.2017.8120647
Thanawit Prasongpongchai, T. Chalidabhongse, Sangsan Leelhapantu
Visual inspection of rail fasteners is crucial to rail safety. However, the traditional method in which railway staffs manually inspect the conditions of fasteners is time-consuming and prone to human error. In this paper, we present a method to automatically detect missing rail fasteners from top-view images. Using a top-down approach, coarse bounding boxes of potential fastener areas are first located from the track and the tie regions with an edge density map and the RANSAC algorithm. Preprocessed with the guided filter, the region within the bounding boxes are then scanned to detect rail fasteners using PHOG features and e-SVR with RBF kernel. The boxes, in which no fasteners are found, are reported as missing fasteners. The proposed method was tested and has shown a degree of robustness in scenes from complex real-world environments with the 100% probability of detection and 3.47% probability of false alarm for missing fastener detection. The results also indicate that the use of guided filter, RBF kernel and the image pyramid technique for feature extraction significantly improves the performance of the classifier.
{"title":"A vision-based method for the detection of missing rail fasteners","authors":"Thanawit Prasongpongchai, T. Chalidabhongse, Sangsan Leelhapantu","doi":"10.1109/ICSIPA.2017.8120647","DOIUrl":"https://doi.org/10.1109/ICSIPA.2017.8120647","url":null,"abstract":"Visual inspection of rail fasteners is crucial to rail safety. However, the traditional method in which railway staffs manually inspect the conditions of fasteners is time-consuming and prone to human error. In this paper, we present a method to automatically detect missing rail fasteners from top-view images. Using a top-down approach, coarse bounding boxes of potential fastener areas are first located from the track and the tie regions with an edge density map and the RANSAC algorithm. Preprocessed with the guided filter, the region within the bounding boxes are then scanned to detect rail fasteners using PHOG features and e-SVR with RBF kernel. The boxes, in which no fasteners are found, are reported as missing fasteners. The proposed method was tested and has shown a degree of robustness in scenes from complex real-world environments with the 100% probability of detection and 3.47% probability of false alarm for missing fastener detection. The results also indicate that the use of guided filter, RBF kernel and the image pyramid technique for feature extraction significantly improves the performance of the classifier.","PeriodicalId":268112,"journal":{"name":"2017 IEEE International Conference on Signal and Image Processing Applications (ICSIPA)","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133869658","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-09-01DOI: 10.1109/ICSIPA.2017.8120641
M. N. Hussien, M. H. Lye, M. F. A. Fauzi, Tan Ching Seong, Sarina Mansor
This paper presents the evaluation of visual features for the proposed two eye detection method applied to thermal images. The use of two eye region is due to its distinctive pattern and to overcome the issue of blurred and noisy characteristic in the thermal image. Comparative performance analysis on three different features which includes Haar, Histogram of Oriented Gradients (HoG) and Local Binary Patterns (LBP) is conducted. The performance of the eyes detection method is measured based on the correct detection of both eyes inside the face image. The experiments were done on the Natural Visible and Infrared Facial Expression Database (NVIE). The method proposed in this paper shows good eye detection accuracy. The best detection accuracy is obtained using the HoG feature.
{"title":"Comparative analysis of eyes detection on face thermal images","authors":"M. N. Hussien, M. H. Lye, M. F. A. Fauzi, Tan Ching Seong, Sarina Mansor","doi":"10.1109/ICSIPA.2017.8120641","DOIUrl":"https://doi.org/10.1109/ICSIPA.2017.8120641","url":null,"abstract":"This paper presents the evaluation of visual features for the proposed two eye detection method applied to thermal images. The use of two eye region is due to its distinctive pattern and to overcome the issue of blurred and noisy characteristic in the thermal image. Comparative performance analysis on three different features which includes Haar, Histogram of Oriented Gradients (HoG) and Local Binary Patterns (LBP) is conducted. The performance of the eyes detection method is measured based on the correct detection of both eyes inside the face image. The experiments were done on the Natural Visible and Infrared Facial Expression Database (NVIE). The method proposed in this paper shows good eye detection accuracy. The best detection accuracy is obtained using the HoG feature.","PeriodicalId":268112,"journal":{"name":"2017 IEEE International Conference on Signal and Image Processing Applications (ICSIPA)","volume":"37 14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131931136","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-09-01DOI: 10.1109/ICSIPA.2017.8120634
P. S. Patil, B. Neole, K. Bhurchandi
Denoising digital images while preserving sharp details and fine edges is an active area of research. This paper presents novel local edge profile detection and preservation based denoising algorithm for digital images in presence of zero mean Gaussian noise. Detecting and preserving sharp changes in image pixel intensities preserves the visual quality of the denoised image. Twenty four different types and orientations of edges are detected in 3×3 overlapping tiles of a picture. Based on the edge type and orientation, each tile is subjected to integration along the direction of the detected edge. This preserves the edges. Simple averaging is done if a tile does not have any edge. Continuity of edges is maintained by taking the overlapping tiles of the same edge and integrating both the neighbouring tiles in the direction of the edge. The integration across the edges are avoided to preserve the sharpness of the edges. The proposed algorithm is benchmarked with other denoising algorithms in terms of a novel edge representation parameter i.e. number of edge tiles in the input image. The proposed algorithm clearly outperforms the other contemporary algorithms. Most of the other algorithms either over construct or under construct the edges during denoising.
{"title":"Performance evaluation of a new local edge profile preservation based denoising algorithm","authors":"P. S. Patil, B. Neole, K. Bhurchandi","doi":"10.1109/ICSIPA.2017.8120634","DOIUrl":"https://doi.org/10.1109/ICSIPA.2017.8120634","url":null,"abstract":"Denoising digital images while preserving sharp details and fine edges is an active area of research. This paper presents novel local edge profile detection and preservation based denoising algorithm for digital images in presence of zero mean Gaussian noise. Detecting and preserving sharp changes in image pixel intensities preserves the visual quality of the denoised image. Twenty four different types and orientations of edges are detected in 3×3 overlapping tiles of a picture. Based on the edge type and orientation, each tile is subjected to integration along the direction of the detected edge. This preserves the edges. Simple averaging is done if a tile does not have any edge. Continuity of edges is maintained by taking the overlapping tiles of the same edge and integrating both the neighbouring tiles in the direction of the edge. The integration across the edges are avoided to preserve the sharpness of the edges. The proposed algorithm is benchmarked with other denoising algorithms in terms of a novel edge representation parameter i.e. number of edge tiles in the input image. The proposed algorithm clearly outperforms the other contemporary algorithms. Most of the other algorithms either over construct or under construct the edges during denoising.","PeriodicalId":268112,"journal":{"name":"2017 IEEE International Conference on Signal and Image Processing Applications (ICSIPA)","volume":"309 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134273073","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-09-01DOI: 10.1109/ICSIPA.2017.8120605
Shubhi Kansal, Shikha Purwar, R. Tripathi
Histogram equalization is the simplest method of image enhancement. Mean brightness and contrast are important parameters of an image. Artifacts are generated if the original mean brightness of an image is not preserved. A high contrast provides good visual quality. Contrast can be increased by increasing entropy of the image. Entropy can be maximized by making the image histogram as flat as possible. In this paper, the trade-off between mean brightness and maximum entropy performed to achieve high contrast image with conserved mean brightness.
{"title":"Trade-off between mean brightness and contrast in histogram equalization technique for image enhancement","authors":"Shubhi Kansal, Shikha Purwar, R. Tripathi","doi":"10.1109/ICSIPA.2017.8120605","DOIUrl":"https://doi.org/10.1109/ICSIPA.2017.8120605","url":null,"abstract":"Histogram equalization is the simplest method of image enhancement. Mean brightness and contrast are important parameters of an image. Artifacts are generated if the original mean brightness of an image is not preserved. A high contrast provides good visual quality. Contrast can be increased by increasing entropy of the image. Entropy can be maximized by making the image histogram as flat as possible. In this paper, the trade-off between mean brightness and maximum entropy performed to achieve high contrast image with conserved mean brightness.","PeriodicalId":268112,"journal":{"name":"2017 IEEE International Conference on Signal and Image Processing Applications (ICSIPA)","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130793478","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-09-01DOI: 10.1109/ICSIPA.2017.8120661
M. Awais, H. Müller, T. Tang, F. Mériaudeau
Diabetic Macular Edema (DME) is one of the many eye diseases that is commonly found in diabetic patients. If it is left untreated it may cause vision loss. This paper focuses on classification of abnormal and normal OCT (Optical Coherence Tomography) image volumes using a pre-trained CNN (Convolutional Neural Network). Using VGG16 (Visual Geometry Group), features are extracted at different layers of the network, e.g. before fully connected layer and after each fully connected layer. On the basis of these features classification was performed using different classifiers and results are higher than recently published work on the same dataset with an accuracy of 87.5%, with sensitivity and specificity being 93.5% and 81% respectively.
{"title":"Classification of SD-OCT images using a Deep learning approach","authors":"M. Awais, H. Müller, T. Tang, F. Mériaudeau","doi":"10.1109/ICSIPA.2017.8120661","DOIUrl":"https://doi.org/10.1109/ICSIPA.2017.8120661","url":null,"abstract":"Diabetic Macular Edema (DME) is one of the many eye diseases that is commonly found in diabetic patients. If it is left untreated it may cause vision loss. This paper focuses on classification of abnormal and normal OCT (Optical Coherence Tomography) image volumes using a pre-trained CNN (Convolutional Neural Network). Using VGG16 (Visual Geometry Group), features are extracted at different layers of the network, e.g. before fully connected layer and after each fully connected layer. On the basis of these features classification was performed using different classifiers and results are higher than recently published work on the same dataset with an accuracy of 87.5%, with sensitivity and specificity being 93.5% and 81% respectively.","PeriodicalId":268112,"journal":{"name":"2017 IEEE International Conference on Signal and Image Processing Applications (ICSIPA)","volume":"94 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122361016","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：