Sharif Digital Repository

In the modern world today, the increasing need for transfering data in high speeds and long distances has encouraged the investors of this area to use more efficient technologies of this field. With regard to the need for extensive and scalable communication networks, reusing the bandwidth in closer distances seems to be inevitable. Hence, finding an optimal way of using resources like bandwidth and power is necessary. In OFDMA-based networks there is no intra-cell interference due to using orthogonal subbands and the most important limiting factor in designing these networks is inter-cell interference. Therefore in this thesis we are gonig to analyase the partial frequency reuse method... 

The labeling process is one of the most important tasks in the field of computer vision. The dense labeling problem is the main step towards 2D and 3D scene understanding. The main goal of dense labeling is to label all pixels of images that are known as a semantic segmentation of images in the related literature. Although the state-of-the-art results are mainly achieved by deep learning methods, traditional methods had also been at the center of attention for some years. In the last decades, convolutional neural networks have changed the landscape of visual recognition tasks such as labeling and semantic segmentation. The most important issues in deep learning models are the hardware and... 

Self-supervied learning is a method to reduce the need for large labeled datasets in supervised learning. In self-supervised learning, the goal is to design a pretext task that can be trained without any labels. This pretext task results in learning a representation of data that can reduce the need for labels when used for different tasks. In the domain of images, data augmenting transformations which are often a composition of simple transformations such as random cropping and color jitter have been used for the design of pretext tasks. These simple transformations can cause information loss in some datasets which limits the usage of the learned representations for various downstream tasks.... 

The remarkable progress of deep neural networks in recent years has led to their entry into the industry and their use in the real world. However, one of the most important and basic issues that threaten the security of these networks is attacks. The attacks that deliberately manipulate input data cause vulnerabilities and misclassify networks. Due to the wide range of ways in which attacks can perturb input data, identifying their types is considered a vital part of ensuring a robust network. The inability of deep networks to generalize to unseen data is also an important limitation. This thesis presents a 2D adversarial attack and a 3D defense in this regard.In 2D attacks, the type of... 

In image processing applications, sensors (Camera, LiDAR and Stereo) are essential for scene perception and Deep learning methods outperform most of the image processing tasks like 3D and 2D object detection and semantic segmentation. Different sensors are used in image processing tasks. Sensor fusion is using multiple sensors data to get better performance. Each sensor captures different data (e.g, color, texture, and depth). Some of them are distorted in inclement weather, intense illuminance changes, and dark environments which multi-sensor data fusion is used to overcome sensor weaknesses. One of the most important fields that sensor fusion used is Auto Driving cars (AD). Different... 

Cardiovascular diseases are one of the leading causes of death worldwide. Despite significant advances in therapeutic interventions for these diseases, the pathophysiology remains incompletely understood. Challenges in drug evaluation include the lack of appropriate clinical trials, such as conventional 2D cell cultures and animal studies, as well as the absence of model systems that can provide complex and patient-specific cardiac functions for diagnosis. Organ-on-a-chip devices have the potential to mimic the body's response to treatments by utilizing microscale cell culture fluid connections and creating realistic models of human organs of interest. Our objective is to develop a... 

The retina, a photosensitive area in the central nervous system, is delicate and intricate. It is susceptible to degenerative disorders such as age-related macular degeneration (AMD), retinitis pigmentosa (RP), Stargardt disease (SD), and glaucoma. These diseases can lead to severe vision loss and ultimately irreversible blindness by causing destruction or dysfunction of different types of retinal cells. Unfortunately, there are no proven treatment strategies to cure or reverse these degenerative disorders. However, cell transplantation therapies may be an alternative to replace distorted cells and improve an individual's vision. Recent clinical outcomes show that transplanted cells in the... 

Cranioplasty is a surgical procedure for repairing skull defects. This surgery will protect the brain tissue, reduce pain in the lesion site and reduce the psychological burden on the patient. Cranioplasty implants should have distinct characteristics, i.e., high strength for protecting the brain, full coverage of skull defects, resistance to infection, non-expansion with heat, and reasonable price. Titanium implants, bone allografts, hydroxyapatite, and methyl methacrylate are commonly used in this surgery. However, these materials have many disadvantages that limit their use. As a result, biodegradable material and 3D printing technology are the next steps for designing scaffolds according... 

Semantic segmentation in computer vision plays a vital role in teaching machines to mimic how humans interpret visual information. In this area, common fully supervised methods face the challenge of manual and time-consuming pixel-level annotations of large datasets. This volume of annotation also places a heavy processing load on the hardware. In recent years, weak supervision has been introduced to address these challenges. In this method, the supervision of the model is done by weak labels, which are more available compared to pixel-wise annotations of an image. This efficiency in the annotation cost and processing load, along with high accuracy, has brought weakly supervised semantic... 

In real-time semantic segmentation, images are divided into predefined semantic regions at a speed exceeding 30 frames per second, and each point of the image is labeled with a category. The output of this fundamental task is to produce a representation that is comprehensible to computers for further processing in applications such as autonomous vehicles, robotics, and augmented reality. This task pursues two conflicting goals: improving the accuracy of segmentation and increasing its speed. Previous research has enhanced segmentation speed using lightweight and multi-branch structures and improved accuracy using attention mechanisms. Depth map alongside color image can lead to better... 

Video Instance Segmentation is one of the newest tasks in computer vision, tasked with segmenting, categorizing, and tracking instances across video frames. This task is highly significant and applicable today in industries such as autonomous vehicles, surveillance systems, production lines, and medical video analysis. Generally, there are two approaches for solving the task of Video Instance Segmentation: the object-oriented approach and the pixel-oriented approach. In the object-oriented approach, after detecting instances at the image level, the segmentation and tracking processes are performed to link the instances. In the pixel-oriented approach, all spatial-temporal information is... 

In this project, we address monocular depth estimation, a fundamental challenge in understanding environmental scenes. Estimating depth from a single image can be particularly challenging due to out-of-distribution information such as occlusions, unexpected objects, and varying lighting conditions, all of which can disrupt model performance. Despite advancements in monocular depth estimation through the development of deep neural networks, challenges remain in accurately estimating depth maps and object boundaries. This research introduces a method aimed at improving the performance and reliability of monocular depth estimation models by leveraging epistemic uncertainty estimation in the... 

Nowadays, with high performance computers and progress in video recorder devices and their related technologies, price of these devices is being balanced and recording the videos is conventional now. Processing of these videos has been a challenge in industry and science. Control and surveillance of places such as airports and roads and reducing their sizes in memory are some of their research area. Usually one of the initial preprocesses of such applications specially the object tracking is background modeling and background subtraction. This preprocess has an essential effect on the other algorithms that process based on these processes’ results. If this section conducted with fewer... 

In recent years, ability of computers to recognize human actions, because of numerousapplications, has attracted scientists. Surveillancesystems in house, work and public places, human computer interaction, study of human movement problems, remote supervision of ill or old people and sport training are only some of the applications. In this thesis 10 actions are considered. These actions are Walking, Running, Galloping side, Bending, Jump jacking, Jumping, Jumping in place, Skipping, Waving one hand and Waving two hands. All actions exist in Weisemann dataset so this dataset is used as training and testing dataset. Here important objectives are recognising human action so that it is... 

In this thesis we will first review the problem of deformable modeling. Due to their corss-disciplinary nature, deformable modeling techniques have been the subject of the vigorous research over the past three decades and have found numerous applications in the field of machine vision. Thus the focus will be on general deformable models for computer-based modeling which can be used for computer graphics, visualization, and various image processing applications. So the state of the art of deformable modeling is discussed. Then the focus of the thesis will be on different approaches for the problem of deformable surface reconstruction in 3D space. Amongs these methods there are two main... 

One major problem in sport match analysis and interpretation is that there are restricted viewpoints of the game. 3D reconstruction allows us to have virtual replay from any viewpoint so that the events such as penalties can be detected easily. For achieving this purpose, we use many cameras with overlapped views that cover almost all parts of the field. After some preprocessing, 3D reconstruction and texture mapping, we can enhance the viewer experience by having arbitrary viewpoint of the match. In this research, we use two different visual hull methods for 3D reconstruction. In the first method, the visual hull of an object is computed efficiently and robustly from image contours. This... 

The goal of human action recognition systems is to label the sensory observation data, using one of the predefined action verbs in the system. During recent years, human action recognition has received a growing interest due to its application in automatic scene interpretation. For instance, automatic surveillance systems in public places should be able to discriminate normal and suspicious actions. Human-computer-interface (HCI) systems (which have became popular in recent years) mostly need a similar system to recognize the gesture of their users without using any keyboard (or similar input devices). Human action recognition technologies can also boost the video retrieval systems.
An... 

Image Registration is a fundamental task in numerous applications in medical image processing and is defined as the process of determining the correspondence of between images collected at different times or using different imaging modalities. This correspondence can be used for aligning images so that the pair can be directly compared, combined or analyzed. Image-guided surgery systems use registration for establishing an accurate relation between preoperative and intraoperative image space. There have been several methods for rigid registration which are not easily applicable for soft tissues. Indeed, nonrigid deformation of soft tissues will endanger the accuracy of rigid methods, so it... 

In many applications of multi view 3D reconstruction, image mosaicing is one of the most important processes. Image mosaicing is the action of stitching some pictures of a scene so that a complete picture of the scene is constructed. The proposed method is such that for PTZ cameras by extracting features from frame, the homography matrix between frames will be computed. Then by converting plane coordinate into spherical coordinate, a complete picture of the scene will be constructed and forground will be removed. For fixed cameras sufficient corresponding points will be selected for computing the homography matrix between cameras. Using the homography, the frames without forground will be... 

In large databases, availability of labeled training data is mostly prohibitive in classification. Semi-supervised algorithms are employed to tackle the lack of labeled training data problem. Video databases are the epitome for such a scenario; that is why semi-supervised learning has found its niche in it. Graph-based methods are a promising platform for semi-supervised video classification. Based on the multiview characteristic of video data, different features have been proposed (such as SIFT, STIP and MFCC) which can be utilized to build a graph. In this project, we have proposed a new classification method which fuses the results of manifold regularization over different graphs. Our...