50 research outputs found
Study Of Shape-Based Semi-Local Point Descriptor For Building Recognition
No full textIn this thesis project, the author has carried out a study involving low-level features and middle-level features that can be used for man-made object recognition. This study focuses on a very specific type of man-made structure, i.e. buildings, taken from image capturing devices from ground-level
CNN-based Occluded Person Re-identification in a Multi Camera Environment
No full textIn the context of rising global urban security concerns and the growing use of surveillance cameras, this study aims to enhance individual identification accuracy in occlusion scenarios using deep learning. Four CNN-based models for person re-identification are analyzed and put into practice. Additionally, comparative studies are conducted, and the model’s performance is assessed using the Market-1501 and Occluded-Reid datasets. We propose the use of ensemble learning and convolutional neural networks (CNNs) to address occlusion issues. Our results show that the ensemble approach performs better in re-identification tasks than traditional deep learning algorithms with an improvement of 1%–2% in mAP and Rank-1 scores, respectively
Content based image retrieval and classification using Speeded-Up Robust Features (SURF) and grouped Bag-of-Visual-Words (BoVW)
No full textThis paper presents a work in progress for a proposed method for Content Based Image Retrieval (CBIR) and Classification. The proposed method makes use of the interest points detector and descriptor called Speeded-Up Robust Features (SURF) combined with Bag-of-Visual-Words (BoVW). The combination yields a good retrieval and classification result when compared to other methods. Moreover, a new dictionary building method in which each group has its own dictionary is also proposed. Our method is tested on the highly diverse COREL1000 database and has shown a more discriminative classification and retrieval result
A Conceptual Approach to Predicting Seismic Events and Flood Risks Using Convolutional Neural Networks
Get PDFThis paper explores the application of convolutional neural networks (CNNs) in predictive modelling for seismic events and flood risks, with a particular focus on forecasting extreme quantile events that exceed historical data limits. Traditional risk assessment methods often struggle to estimate such extremes, highlighting the need for more advanced predictive models capable of handling rare but high-impact events. This research enhances CNN architecture to improve accuracy in high quantile predictions by integrating multi-source spatiotemporal data, addressing a critical research gap. The methodology involves incorporating diverse datasets, including geospatial, meteorological, and historical seismic or flood records, into CNN models to augment predictive capabilities. These models undergo systematic validation using historical events and real-world data to assess their reliability, robustness, and practical relevance. Furthermore, the study evaluates the potential of these advanced prediction models to inform disaster risk management and mitigation strategies. By leveraging deep learning techniques and optimizing CNN structures, this research aims to refine forecasting precision, supporting proactive disaster preparedness. The anticipated outcome is an improved predictive framework that enhances early warning systems, facilitates informed decision-making, and strengthens emergency response mechanisms. Ultimately, this study contributes to the broader goal of increasing resilience against natural disasters by equipping policymakers, emergency responders, and urban planners with more accurate and timely risk assessments
Vision-based hand gesture recognition from RGB video data using SVM
No full textWith the growing number of population in the world nowadays, novel human-computer interaction systems and techniques can be used to help improve their quality of life. A gesture based technology can help to maintain the safety and needs of the disable as well as the general people. Gesture recognition from video streams is a challenging task due to the high changeability in the features of each gesture with respect to different person. In this work, we propose a vision-based hand gesture recognition from RGB video data using SVM. Gesture-based interfaces are more natural, spontaneous, and straightforward. Previous works attempted to recognize hand gesture for different scenarios. Throughout our studies, gesture recognition system can be based on wearable sensor or it can be vision based. Our proposed method is applied on a vision based gesture recognition system. In our proposed system image acquisition starts from RGB videos capture using Kinect sensor. We convert the image frames from videos to blur for background noise removal. Then, we convert the images into hsv color mode. After that, we do the dilation, erosion, filtering, and thresholding the image for converting to black and white format. Finally, using the prominent classification algorithm SVM, hand gestures has been recognized. In conclusion, the framework aims to create a better vision-based hand gesture recognition system with novel techniques.
© (2019) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only
Simplified Local Binary Pattern Orientation Histogram for Building Image Retrieval
No full textInternational audienc
Vision Based Gesture Recognition from RGB Video frames Using Morphological Image Processing Techniques
No full textwith the large number of population in all over the world nowadays, novel human computer interaction systems and techniques can be used to help improve our way of life.
A vision based gesture recognition technology can help to maintain the safety and needs of the disable as well as others. Gesture recognition from video frames is a challenging
task due to the high changeability in the features of each gesture with respect to different person. In this work, we propose a vision-based hand gesture recognition algorithm where the image frames are from RGB video data. Gesture-based systems are more
natural, spontaneous, and straightforward. Previous works attempted to recognize hand gesture for different kind of scenarios. According to our studies, gesture recognition
system can be based on wearable sensor or it can be vision based. Our proposed method is applied on a vision based gesture recognition system. In our proposed system image acquisition starts from RGB videos capture using Kinect sensor. We convert the image frames one after another from videos to blur for background noise removal. Then, we convert the images of a whole video into HSV color mode. After that, we do the dilation, erosion, filtering, and thresholding operations on the images. We use these morphological image processing techniques for converting the images to black and white format. Finally, using the prominent classification algorithm SVM we recognize the hand gestures
with a higher accuracy 91.01 percent compared to the state of the art. In conclusion, the proposed algorithm aims to create a better vision-based hand gesture recognition system with a unique solution in this domain
Combining Local Binary Pattern and Edge Orientation for Shape Based Image Retrieval
No full textInternational audienc
Combinatorial Filtering and Machine Learning Approach to Improve Vessel Segmentation for Retinal Image Analysis
No full textIn assessing prevalent eye-diseases such Glaucoma,
Diabetic Retinopathy (DR), Age-related macular
degeneration (AMD) etc. retinal image understanding is
crucial [1],[2].
• Fundus cameras are used to capture the fundus images
which visualize the interior surface of the eye.
• Fundus images are analyzed by the ophthalmologist to
identify any abnormalities within the retinal structures.
• Retinal blood vessel (RBV) and Optic Disc (OD) structures
found in fundus images are segmented and localized to
diagnose diseases
