Jharna Majumdar | Nitte Meenakshi Institute Of Technology (original) (raw)
Papers by Jharna Majumdar
Journal of Engineering Research and Reports, 2021
Image Fusion is a process of adding information obtained from various sensors and intelligent sys... more Image Fusion is a process of adding information obtained from various sensors and intelligent systems. This provides an image which containing complete information. In this process we fuse images of the same scene one is Infrared Image and other is visible image to produce an image that contains more information. In this the Infrared which are low resolution and noisy nature of image and visual image fused using Stationary wavelet transfer algorithm. In this we have used Gabor filter and GLCM to extract feature and compared the two feature extraction method using quality matrix parameters and found which method is the best method of fusion. The fusion is used in satellite image fusion which mostly used in maps also for decision-making process involved in interpreting images from multi sensor data. Images of several different targets (a military vehicle, a wood chipper, a pickup truck, and people) were used to assess how human subjects view and interpret different types of images. Th...
IOSR Journal of Computer Engineering, 2014
Video summarization is a process of removing the redundant frames and generating the most informa... more Video summarization is a process of removing the redundant frames and generating the most informative key frames of the videos. In this paper we have explained two efficient methods for video summarization and given the comparison between these two methods. The first method is Video summarization using CLD feature extraction and adaptive threshold technique for shot boundary detection. Second method is video summarization using aggregation function where three different features such as Histogram difference, correlation difference and moment of inertia difference are combined and that difference value is compared with the predefined threshold value.
International Journal of Engineering Research and
Automatic Human detection and tracking is a vital part of video surveillance. Many human detectio... more Automatic Human detection and tracking is a vital part of video surveillance. Many human detection and human tracking algorithms have been discussed in literature survey. Authors in this paper have attempted to identify the human in clattered environment, identify human body (head, body and leg, track the human in the video based on RGB colour model and also detect collision between multiple human.
International Journal of Computer and Communication Technology, 2011
Shape and characteristics of the histogram plays a major role in finding the quality of an image.... more Shape and characteristics of the histogram plays a major role in finding the quality of an image. Histogram Specification is an image enhancement technique, where the histogram of the input image is transformed to a pre-specified histogram derived from a high resolution image, called target image. In this paper, the classical histogram specification technique is extended by using a target image which is obtained by fusing multiple high resolution images. A set of Quality Metrics were identified to assess the quality of the output enhanced image. The paper addresses the following issues: a) Effect of varying the number of target images on the quality of the output enhanced image b) Role of using different methods of fusion on the quality of the output enhanced image c) Category of the target image on the quality of the output enhanced image. If the input image is from a forest, whether in order to obtain an enhanced image, all target images has to be selected from the forest category...
Emerging Research in Computing, Information, Communication and Applications, 2019
In digitized world, data is growing exponentially and Big Data Analytics is an emerging trend and... more In digitized world, data is growing exponentially and Big Data Analytics is an emerging trend and a dominant research field. Data mining techniques play an energetic role in the application of Big Data in healthcare sector. Data mining algorithms give an exposure to analyse, detect and predict the presence of disease and help doctors in decision-making by early detection and right management. The main objective of data mining techniques in healthcare systems is to design an automated tool which diagnoses the medical data and intimates the patients and doctors about the intensity of the disease and the type of treatment to be best practiced based on the symptoms, patient record and treatment history. This paper emphasises on diabetes medical data where classification and clustering algorithms are implemented and the efficiency of the same is examined.
International Journal of Computer and Communication Technology, 2011
Image enhancement has been an area of active research for decades. Most of the studies are aimed ... more Image enhancement has been an area of active research for decades. Most of the studies are aimed at improving the quality of image for better visualization. An approach for contrast enhancement utilizing multi-scale analysis is introduced. To show the effects of image enhancement, quantitative measures should be introduced. In this paper, we examine the effect of global and local enhancement using multi resolution pyramids. We identify a set of quality metric parameters for comparative performance analysis and use it to assess the enhanced output image for a number of image enhancement algorithms using pyramids.
Advances in Intelligent Systems and Computing, 2018
Optical Flow is the apparent motion of pixels that is generated when there is relative motion bet... more Optical Flow is the apparent motion of pixels that is generated when there is relative motion between an observer and a scene. Optical flow is used in many areas of research for object detection, motion estimation, navigation and tracking applications. In this paper, we have proposed two novel applications where Optical Flow has been used for Determining Shot Transitions in a video sequence and Human-Face Expression Detection in a video. For Video Shot Detection, local invariant feature points using SIFT (Scale Invariant Feature Transform) corner detectors is calculated and Optical Flow is computed with the detected feature points to determine shot changes in the video. Quality Parameters like Recall, Precision and F-measure is used to determine the quality of the algorithm. Whereas for detecting of human faces, we begin by first performing skin segmentation to obtain probable regions where human face is present. Within the probable region Optical Flow is used to eliminate the background and other objects having human skin color. From the isolated face, Optical Flow is used for identifying expressions.
Emerging Research in Computing, Information, Communication and Applications, 2021
Proceedings of the Third International Conference on Advanced Informatics for Computing Research, 2019
Retinex enhancement is an extremely powerful algorithm in the domain of image/video enhancement d... more Retinex enhancement is an extremely powerful algorithm in the domain of image/video enhancement due to its ability to deliver effective enhancement It differs from the regular enhancement under very bad illumination, as well as, bad weather conditions. However, its algorithmic complexity is very high, therefore, it cannot be used for real-time enhancement. This paper presents the work done on accelerating performance of the Retinex algorithms, SSR and MSR, using multithreading software optimization and hardware optimization, on the embedded platforms- UDOO x86 Ultra and Nvidia Jetson Tegra K1 and also the integration of these boards with a mobile robot, for tracking in cluttered environment, where the velocity control of the robot is achieved using MPC.
2018 International Conference on Advances in Computing, Communications and Informatics (ICACCI), 2018
Video Shot Detection is utilized to detect change of scene in the video. Video Summarization give... more Video Shot Detection is utilized to detect change of scene in the video. Video Summarization gives most useful frames to create the conceptual information of the given video. In this paper, we have coordinated both the ideas of shot detection and summarization for the better result of the outcomes. The SIFT features are extracted from the frames and compared with consecutive frames for Video Shot detection and the Video Summarization is done using Expected-Maximization. The shot frames of Video shot detection are utilized as contribution for the Video Summarization. The quality metric parameters are utilized to assess the strength of the key frame extraction.
2017 3rd International Conference on Control, Automation and Robotics (ICCAR), 2017
Design and Development of an Android (A robot with human resemblance and behaviour) is complex wh... more Design and Development of an Android (A robot with human resemblance and behaviour) is complex which requires high level of dexterity and cognitive capability. In the present work, we propose a low cost minimalist mechanical design of an android robot called ARIA (Advanced Robot for Interactive Applications) which has 12 Degrees of Freedom (DoF). ARIA is capable of expressing various emotions such as Sadness, Surprise, Fear, Anger, and Tiredness etc, along with various neck articulations such as flexion, extension, lateral flexion and rotation. In the present work, we have adopted the minimalist design approach to design the android with an overall objective of reducing the design complexity and development time. We have described in detail the basic rationale behind minimalist design approach, the design of facial features, its kinematic linkage and actuation mechanisms; and finally we have conducted the experiments to characterize and validate the effectiveness of the facial expressions of the minimalist android. The experimental results reveal that the design trade-offs as a result of minimalist mechanical design highly impacts the interactive dynamics between humans and robot. The cognitive perceptions of humans towards androids demand more realistic facial expressions which were not possible to achieve using minimally designed ARIA.
ICTACT Journal on Image and Video Processing, 2017
This paper provides an early attempt to train and retrieve handwritten Nandinagari characters usi... more This paper provides an early attempt to train and retrieve handwritten Nandinagari characters using one of the latest techniques in visual feature detection. The data set consists of over 1600 handwritten Nandinagari characters of different fonts, size, rotation, translation and image formats. In the Learning phase, we subject them to an approach where their recognition is effective by first extracting their key interest points on the images which are invariant to Scale, rotation, translation, illumination and occlusion. The technique used for this phase is Scale Invariant Feature Transform (SIFT). These features are represented in quantized form as visual words in code book generation step. Then the Vector of Locally Aggregated Descriptors (VLAD) is used for encoding each of the Image descriptors in the database. In the recognition phase, for query image, SIFT features are extracted and represented as query vector .Then these features are compared against the visual vocabulary generated by code book to retrieve similar images from the database. The performance is analysed by computing mean average precision .This is a novel scalable approach for recognition of rare handwritten Nandinagari characters with about 98% search accuracy with a good efficiency and relatively low memory usage requirements.
Image registration is an important process in high-level image interpretation systems developed f... more Image registration is an important process in high-level image interpretation systems developed for civilian and defense applications. Registration is carried out in two phases : namely feature extraction and feature correspondence. The basic building block of feature based image registration scheme involves matching feature points that are extracted from a sensed image to their counter parts in the reference image. Features may be control points, corners, junctions or interest points. The objective of this study is to develop a methodology for iterative convergence of transformation parameters automatically between two successive pairs of aerial or satellite images. In this paper we propose an iterative image registration approach to compute accurate and stable transformation parameters that would handle image sequences acquired under varying environmental conditions. The iterative registration procedure was initially tested using satellite images with known transformation paramete...
Advances in Intelligent Systems and Computing, 2018
With rapid development in Digital Video Technology capturing of video has become very easy and an... more With rapid development in Digital Video Technology capturing of video has become very easy and an integral part of our lives. Exponentially growing and already existing video data must be managed effectively. This paper, which is an extension of our previous research work, explores new distribution domains in addition to the distributions discussed in our earlier work in order to determine the best one for the purpose of detection of ‘cut’ transitions in videos. Simple shot detection algorithms have been used for finding the “cut” transition. Appropriate memory models have been used so that the algorithm runs seamlessly on the video. Finally, from the analysis of results we find out the distribution and video shot detection model which gives the best accuracy and efficiency.
Defence Science Journal, 2008
A wide variety of image processing applications require segmentation and classification ofimages.... more A wide variety of image processing applications require segmentation and classification ofimages. The problem becomes complex when the images are obtained in an uncontrolledenvironment with a non-uniform illumination. The selection of suitable features is a critical partof an image segmentation and classification process, where the basic objective is to identify theimage regions that are homogeneous but dissimilar to all spatially adjacent regions. This paperproposes an automatic method for the classification of a terrain using image features such asintensity, texture, and edge. The textural features are calculated using statistics of geometricalattributes of connected regions in a sequence of binary images obtained from a texture image.A pixel-wise image segmentation scheme using a multi-resolution pyramid is used to correct thesegmentation process so as to get homogeneous image regions. Localisation of texture boundariesis done using a refined-edge map obtained by convolution, thi...
Advanced Science Letters, 2017
2018 3rd International Conference on Circuits, Control, Communication and Computing (I4C), 2018
Target tracking at a less illuminated environment is a tedious work. In this paper, Adaptive Enha... more Target tracking at a less illuminated environment is a tedious work. In this paper, Adaptive Enhancement algorithms are used as a pre-processing technique. The performance of these methods on real-time scenario has been tested and using Quality Metrix Parameter their usefulness is proved. Particle Filter Based Tracking [5] is used for the purpose of Target/Human Tracking. A mobile robot is used for tracking in low illumination. A robust Control system is required for the robot to efficiently follow the human using the computer vision algorithm. Hence, Linear Quadratic Integral (LQI) has been implemented on the system for velocity control. The velocity of the robot is monitored such that it doesn’t crash into the target.
Proceedings of the Third International Conference on Advanced Informatics for Computing Research, 2019
Visual object Tracking is one of the most challenging tasks in computer vision due to various com... more Visual object Tracking is one of the most challenging tasks in computer vision due to various complications like environmental clutter and object clutter. In this paper we propose the use of Masked RCNN and YoloV2 based CNN architecture to overcome the challenges of tracking and we have also compared their performance in real-time application on a Mobile Robot. The type of dataset required, and approach considered for each of the approach to increase the accuracy as well as implantability on real-time system is also discussed. A Skid Steer Mobile Robot (SSMR) is used to follow the human detected by the CNN algorithms. Te Robot Control is done by use of Linear Quadratic Gaussian Controller for velocity control.
In the current situation, COVID19 is one of the life-threatening respiratory infections which inf... more In the current situation, COVID19 is one of the life-threatening respiratory infections which infect humans as well as animal species. The early and accurate detection of COVID19 is essential to make proper decisions and to ensure recovery treatment for patients which will help to save patients lives. Deep Learning approaches are successfully used to analyse and detect COVID19 on chest X-ray and CT scan images. In this paper, a semi supervised learning approach is used to segment the covid affected region, using DeepLabV3 from Chest X-ray (CXR), and the ground truths for the segmentation is created by pre-processing the output class activation maps of DenseNet201, which is used for multiclass classification of COVID19 and non COVID19 X-rays. The performance of the model is evaluated by varying the optimization function, number of training epochs, scheduling learning rates.
The Process of shot boundary detection is a fundamental requirement in automatic video indexing, ... more The Process of shot boundary detection is a fundamental requirement in automatic video indexing, editing and archiving. Many algorithms have been proposed for detecting video shot boundaries and classifying shot and shot transition types. This paper presents a comparison of several new shot boundary detection and classification techniques and their variations including Histograms, Discrete wavelet transform, Haar wavelet based video shot detection and VGRAPH Methods. The performance and ease of selecting good thresholds for these algorithms are evaluated based on a wide variety of video sequences. Threshold selection requires a trade-off between recall and precision that must be guided by the target application.
Journal of Engineering Research and Reports, 2021
Image Fusion is a process of adding information obtained from various sensors and intelligent sys... more Image Fusion is a process of adding information obtained from various sensors and intelligent systems. This provides an image which containing complete information. In this process we fuse images of the same scene one is Infrared Image and other is visible image to produce an image that contains more information. In this the Infrared which are low resolution and noisy nature of image and visual image fused using Stationary wavelet transfer algorithm. In this we have used Gabor filter and GLCM to extract feature and compared the two feature extraction method using quality matrix parameters and found which method is the best method of fusion. The fusion is used in satellite image fusion which mostly used in maps also for decision-making process involved in interpreting images from multi sensor data. Images of several different targets (a military vehicle, a wood chipper, a pickup truck, and people) were used to assess how human subjects view and interpret different types of images. Th...
IOSR Journal of Computer Engineering, 2014
Video summarization is a process of removing the redundant frames and generating the most informa... more Video summarization is a process of removing the redundant frames and generating the most informative key frames of the videos. In this paper we have explained two efficient methods for video summarization and given the comparison between these two methods. The first method is Video summarization using CLD feature extraction and adaptive threshold technique for shot boundary detection. Second method is video summarization using aggregation function where three different features such as Histogram difference, correlation difference and moment of inertia difference are combined and that difference value is compared with the predefined threshold value.
International Journal of Engineering Research and
Automatic Human detection and tracking is a vital part of video surveillance. Many human detectio... more Automatic Human detection and tracking is a vital part of video surveillance. Many human detection and human tracking algorithms have been discussed in literature survey. Authors in this paper have attempted to identify the human in clattered environment, identify human body (head, body and leg, track the human in the video based on RGB colour model and also detect collision between multiple human.
International Journal of Computer and Communication Technology, 2011
Shape and characteristics of the histogram plays a major role in finding the quality of an image.... more Shape and characteristics of the histogram plays a major role in finding the quality of an image. Histogram Specification is an image enhancement technique, where the histogram of the input image is transformed to a pre-specified histogram derived from a high resolution image, called target image. In this paper, the classical histogram specification technique is extended by using a target image which is obtained by fusing multiple high resolution images. A set of Quality Metrics were identified to assess the quality of the output enhanced image. The paper addresses the following issues: a) Effect of varying the number of target images on the quality of the output enhanced image b) Role of using different methods of fusion on the quality of the output enhanced image c) Category of the target image on the quality of the output enhanced image. If the input image is from a forest, whether in order to obtain an enhanced image, all target images has to be selected from the forest category...
Emerging Research in Computing, Information, Communication and Applications, 2019
In digitized world, data is growing exponentially and Big Data Analytics is an emerging trend and... more In digitized world, data is growing exponentially and Big Data Analytics is an emerging trend and a dominant research field. Data mining techniques play an energetic role in the application of Big Data in healthcare sector. Data mining algorithms give an exposure to analyse, detect and predict the presence of disease and help doctors in decision-making by early detection and right management. The main objective of data mining techniques in healthcare systems is to design an automated tool which diagnoses the medical data and intimates the patients and doctors about the intensity of the disease and the type of treatment to be best practiced based on the symptoms, patient record and treatment history. This paper emphasises on diabetes medical data where classification and clustering algorithms are implemented and the efficiency of the same is examined.
International Journal of Computer and Communication Technology, 2011
Image enhancement has been an area of active research for decades. Most of the studies are aimed ... more Image enhancement has been an area of active research for decades. Most of the studies are aimed at improving the quality of image for better visualization. An approach for contrast enhancement utilizing multi-scale analysis is introduced. To show the effects of image enhancement, quantitative measures should be introduced. In this paper, we examine the effect of global and local enhancement using multi resolution pyramids. We identify a set of quality metric parameters for comparative performance analysis and use it to assess the enhanced output image for a number of image enhancement algorithms using pyramids.
Advances in Intelligent Systems and Computing, 2018
Optical Flow is the apparent motion of pixels that is generated when there is relative motion bet... more Optical Flow is the apparent motion of pixels that is generated when there is relative motion between an observer and a scene. Optical flow is used in many areas of research for object detection, motion estimation, navigation and tracking applications. In this paper, we have proposed two novel applications where Optical Flow has been used for Determining Shot Transitions in a video sequence and Human-Face Expression Detection in a video. For Video Shot Detection, local invariant feature points using SIFT (Scale Invariant Feature Transform) corner detectors is calculated and Optical Flow is computed with the detected feature points to determine shot changes in the video. Quality Parameters like Recall, Precision and F-measure is used to determine the quality of the algorithm. Whereas for detecting of human faces, we begin by first performing skin segmentation to obtain probable regions where human face is present. Within the probable region Optical Flow is used to eliminate the background and other objects having human skin color. From the isolated face, Optical Flow is used for identifying expressions.
Emerging Research in Computing, Information, Communication and Applications, 2021
Proceedings of the Third International Conference on Advanced Informatics for Computing Research, 2019
Retinex enhancement is an extremely powerful algorithm in the domain of image/video enhancement d... more Retinex enhancement is an extremely powerful algorithm in the domain of image/video enhancement due to its ability to deliver effective enhancement It differs from the regular enhancement under very bad illumination, as well as, bad weather conditions. However, its algorithmic complexity is very high, therefore, it cannot be used for real-time enhancement. This paper presents the work done on accelerating performance of the Retinex algorithms, SSR and MSR, using multithreading software optimization and hardware optimization, on the embedded platforms- UDOO x86 Ultra and Nvidia Jetson Tegra K1 and also the integration of these boards with a mobile robot, for tracking in cluttered environment, where the velocity control of the robot is achieved using MPC.
2018 International Conference on Advances in Computing, Communications and Informatics (ICACCI), 2018
Video Shot Detection is utilized to detect change of scene in the video. Video Summarization give... more Video Shot Detection is utilized to detect change of scene in the video. Video Summarization gives most useful frames to create the conceptual information of the given video. In this paper, we have coordinated both the ideas of shot detection and summarization for the better result of the outcomes. The SIFT features are extracted from the frames and compared with consecutive frames for Video Shot detection and the Video Summarization is done using Expected-Maximization. The shot frames of Video shot detection are utilized as contribution for the Video Summarization. The quality metric parameters are utilized to assess the strength of the key frame extraction.
2017 3rd International Conference on Control, Automation and Robotics (ICCAR), 2017
Design and Development of an Android (A robot with human resemblance and behaviour) is complex wh... more Design and Development of an Android (A robot with human resemblance and behaviour) is complex which requires high level of dexterity and cognitive capability. In the present work, we propose a low cost minimalist mechanical design of an android robot called ARIA (Advanced Robot for Interactive Applications) which has 12 Degrees of Freedom (DoF). ARIA is capable of expressing various emotions such as Sadness, Surprise, Fear, Anger, and Tiredness etc, along with various neck articulations such as flexion, extension, lateral flexion and rotation. In the present work, we have adopted the minimalist design approach to design the android with an overall objective of reducing the design complexity and development time. We have described in detail the basic rationale behind minimalist design approach, the design of facial features, its kinematic linkage and actuation mechanisms; and finally we have conducted the experiments to characterize and validate the effectiveness of the facial expressions of the minimalist android. The experimental results reveal that the design trade-offs as a result of minimalist mechanical design highly impacts the interactive dynamics between humans and robot. The cognitive perceptions of humans towards androids demand more realistic facial expressions which were not possible to achieve using minimally designed ARIA.
ICTACT Journal on Image and Video Processing, 2017
This paper provides an early attempt to train and retrieve handwritten Nandinagari characters usi... more This paper provides an early attempt to train and retrieve handwritten Nandinagari characters using one of the latest techniques in visual feature detection. The data set consists of over 1600 handwritten Nandinagari characters of different fonts, size, rotation, translation and image formats. In the Learning phase, we subject them to an approach where their recognition is effective by first extracting their key interest points on the images which are invariant to Scale, rotation, translation, illumination and occlusion. The technique used for this phase is Scale Invariant Feature Transform (SIFT). These features are represented in quantized form as visual words in code book generation step. Then the Vector of Locally Aggregated Descriptors (VLAD) is used for encoding each of the Image descriptors in the database. In the recognition phase, for query image, SIFT features are extracted and represented as query vector .Then these features are compared against the visual vocabulary generated by code book to retrieve similar images from the database. The performance is analysed by computing mean average precision .This is a novel scalable approach for recognition of rare handwritten Nandinagari characters with about 98% search accuracy with a good efficiency and relatively low memory usage requirements.
Image registration is an important process in high-level image interpretation systems developed f... more Image registration is an important process in high-level image interpretation systems developed for civilian and defense applications. Registration is carried out in two phases : namely feature extraction and feature correspondence. The basic building block of feature based image registration scheme involves matching feature points that are extracted from a sensed image to their counter parts in the reference image. Features may be control points, corners, junctions or interest points. The objective of this study is to develop a methodology for iterative convergence of transformation parameters automatically between two successive pairs of aerial or satellite images. In this paper we propose an iterative image registration approach to compute accurate and stable transformation parameters that would handle image sequences acquired under varying environmental conditions. The iterative registration procedure was initially tested using satellite images with known transformation paramete...
Advances in Intelligent Systems and Computing, 2018
With rapid development in Digital Video Technology capturing of video has become very easy and an... more With rapid development in Digital Video Technology capturing of video has become very easy and an integral part of our lives. Exponentially growing and already existing video data must be managed effectively. This paper, which is an extension of our previous research work, explores new distribution domains in addition to the distributions discussed in our earlier work in order to determine the best one for the purpose of detection of ‘cut’ transitions in videos. Simple shot detection algorithms have been used for finding the “cut” transition. Appropriate memory models have been used so that the algorithm runs seamlessly on the video. Finally, from the analysis of results we find out the distribution and video shot detection model which gives the best accuracy and efficiency.
Defence Science Journal, 2008
A wide variety of image processing applications require segmentation and classification ofimages.... more A wide variety of image processing applications require segmentation and classification ofimages. The problem becomes complex when the images are obtained in an uncontrolledenvironment with a non-uniform illumination. The selection of suitable features is a critical partof an image segmentation and classification process, where the basic objective is to identify theimage regions that are homogeneous but dissimilar to all spatially adjacent regions. This paperproposes an automatic method for the classification of a terrain using image features such asintensity, texture, and edge. The textural features are calculated using statistics of geometricalattributes of connected regions in a sequence of binary images obtained from a texture image.A pixel-wise image segmentation scheme using a multi-resolution pyramid is used to correct thesegmentation process so as to get homogeneous image regions. Localisation of texture boundariesis done using a refined-edge map obtained by convolution, thi...
Advanced Science Letters, 2017
2018 3rd International Conference on Circuits, Control, Communication and Computing (I4C), 2018
Target tracking at a less illuminated environment is a tedious work. In this paper, Adaptive Enha... more Target tracking at a less illuminated environment is a tedious work. In this paper, Adaptive Enhancement algorithms are used as a pre-processing technique. The performance of these methods on real-time scenario has been tested and using Quality Metrix Parameter their usefulness is proved. Particle Filter Based Tracking [5] is used for the purpose of Target/Human Tracking. A mobile robot is used for tracking in low illumination. A robust Control system is required for the robot to efficiently follow the human using the computer vision algorithm. Hence, Linear Quadratic Integral (LQI) has been implemented on the system for velocity control. The velocity of the robot is monitored such that it doesn’t crash into the target.
Proceedings of the Third International Conference on Advanced Informatics for Computing Research, 2019
Visual object Tracking is one of the most challenging tasks in computer vision due to various com... more Visual object Tracking is one of the most challenging tasks in computer vision due to various complications like environmental clutter and object clutter. In this paper we propose the use of Masked RCNN and YoloV2 based CNN architecture to overcome the challenges of tracking and we have also compared their performance in real-time application on a Mobile Robot. The type of dataset required, and approach considered for each of the approach to increase the accuracy as well as implantability on real-time system is also discussed. A Skid Steer Mobile Robot (SSMR) is used to follow the human detected by the CNN algorithms. Te Robot Control is done by use of Linear Quadratic Gaussian Controller for velocity control.
In the current situation, COVID19 is one of the life-threatening respiratory infections which inf... more In the current situation, COVID19 is one of the life-threatening respiratory infections which infect humans as well as animal species. The early and accurate detection of COVID19 is essential to make proper decisions and to ensure recovery treatment for patients which will help to save patients lives. Deep Learning approaches are successfully used to analyse and detect COVID19 on chest X-ray and CT scan images. In this paper, a semi supervised learning approach is used to segment the covid affected region, using DeepLabV3 from Chest X-ray (CXR), and the ground truths for the segmentation is created by pre-processing the output class activation maps of DenseNet201, which is used for multiclass classification of COVID19 and non COVID19 X-rays. The performance of the model is evaluated by varying the optimization function, number of training epochs, scheduling learning rates.
The Process of shot boundary detection is a fundamental requirement in automatic video indexing, ... more The Process of shot boundary detection is a fundamental requirement in automatic video indexing, editing and archiving. Many algorithms have been proposed for detecting video shot boundaries and classifying shot and shot transition types. This paper presents a comparison of several new shot boundary detection and classification techniques and their variations including Histograms, Discrete wavelet transform, Haar wavelet based video shot detection and VGRAPH Methods. The performance and ease of selecting good thresholds for these algorithms are evaluated based on a wide variety of video sequences. Threshold selection requires a trade-off between recall and precision that must be guided by the target application.