Sharif Digital Repository

To overcome challenges such as object appearance changes and environment illumination variations in tracking methods, online algorithms are suggested to be used instead of offline ones. Online algorithms update the model by the information acquired in the last processed frame. The main challenge of using online algorithms is the accumulation of small errors after several steps of updating of the model (drift) which disturbs the model and causes tracking failure. Using the object information in the first frame in each update can be considered as a solution. The proposed online semi-supervised boosting algorithms can overcome the drift problem at the expense of decreasing their capabilities in... 

With increasing use of computers in everyday life, improved communication between machines and human is needed. To make a right communication and understand a humankind face which is made in a graphical environment, implementing the audio and visual projects like lip reading, audio and visual speech recognition and lip modelling needed. The main goal in this project is natural representation of strings of lip movements for Farsi language. Lack of a complete audio and visual database for this application in Farsi language made us provide a new complete Farsi database for this project that is called SFAVD. It is a unique audio and visual database which covers the most applicable words, all... 

Visual tracking is a classic problem, but is continuously an active area of research, in computer vision. Given a bounding box defining the object of interest (target) in the first frame of a video sequence, the goal of a general tracker is to determine the ob-ject’s bounding box in subsequent frames. Utilizing sparse representation, we propose a robust tracking algorithm to handle challenges such as illumination variation, pose change, and occlusion. Object appearance is modeled using a dictionary composed of target patch images contained in previous frames. In each frame, the target is found from a set of candidates via a likelihood measure that is proportional to the sum of the... 

Bag of words-based image classification systems have achieved state-of-the-art accuracies in the image classification task recently. These systems can be decomposed into four separate subsystems, each of which has its own objectives: Feature extraction, Feature learning and coding, Pooling, and classification. The effects of the feature learning stage, in which each extracted feature is represented as a linear combination of several visual words, can not be neglected in the success of the whole system. The importance of this part has attracted several researchers to develop different methods in order to alleviate the existing issues. Although several methods have been proposed so far, there... 

Tracking is one of the old and still not thoroughly solved problems in machine vision. Its importance lies on its many applications. These applications vary from security surveillance to examining the motion pattern of atomic particles. There is not a tracker which has acceptable results in all situations, yet. A tracker faces many difficulties such as change in illumination and occlusion. In past, tracking was done by using filters or optical flows. By use of the advances in machine learning and pattern recognition, many models have been proposed to accomplish tracking by using these new learning methods. In this dissertation, we proposed a new tracking method which utilizes sparse... 

Due to the recent advances in the learning of multimodal data, humans tend to use computer systems in order to solve more complex problems. One of them is Visual Question Answering (VQA), where the goal is finding the answer of a question asked about the visual contents of a given image. This is an interdisciplinary problem between the areas of Computer Vision, Natural Language Processing and Reasoning. Because of recent achievements of Deep Neural Networks in these areas, recent works used them to address the VQA task. In this thesis, three different methods have been proposed which adding each of them to existing solutions to the VQA problem can improve their results. First method tries to... 

Attention is a cognitive process in which the mind reacts to certain stimuli or stimuli of the environment while other environmental stimuli are ignored. Attention might be an overt or covert process. Overt attention is a process in which based on the purpose, we selectively choose an object or place among other objects and places to focus on and we are aware of it. However, the covert attention originates from hidden source, and we are not aware of it. In fact, the covert attention causes a clear and rapid movement of the eye toward the stimulus or space to be taken into consideration and the time when the movement of the eye it means overt attention has occurred. Visual attention is given... 

The brain-computer interface is an advanced technology in human-machine interaction. The Speller system is a typical use of BCI, in which the target stimulation is detected by the induced signal in the brain. The most commonly speller system, the matrix Speller, has a major disadvantage, and it is Gaze-dependent. Research has proven that target-character selection in the matrix Speller is dependent on eye movement, or as referred to in technical terminology, it is gaze dependent. Therefore, the Speller matrix is not usable for users suffering from unimpaired oculomotor control. Many researchers attempted to overcome this issue, and their results led to two solutions; 1) changing the type of... 

The importance of having a rich repertoire of vocabulary at one’s disposal is evidently an undeniable fact for a successful communication. However, the best way to improve the lexical knowledge of the students is an issue that is still open to dispute. Considering this issue the present study focused on vocabulary learning strategies in L2 teaching. Actually, this study aimed at investigating three different methods of teaching new words to Iranian students in order to determine which one of the methods is more effective in terms of immediate and delayed retention. So the researchers strived to examine the question of how two different mnemonic devices, i.e., the Keyword method and the... 

Melanoma is a malignant skin cancer which is caused by cancerous growth of melanocytes. If not treated at its early development stages, melanoma is the deadliest form of cancer. The best way to cure melanoma is to treat it in its earliest stage of development. Since a melanoma leasion is similar to benign moles (regaring its shape and appearance) at its early stages of development, it is often mistaken for moles and left untreated. Therefore, automatic melanoma detection can increase the survival rate of patients by detecting melanoma in its early stages. In this thesis, a new method for automatic diagnosis of melanoma using segmented dermoscopic images is provided. Almost all related... 

Simultaneous localization and mapping (SLAM) is the action of detecting robot pose in an unknown environment and building the environment map by use of input data that captured from robot sensors. In visual SLAM, the input data for the robot, is limited to camera sensors. Nowadays, SLAM is one of the main challenges in robotic research.For autonomous action, we need robot pose in the map of the environment. The map production in the indoor environment, there is no GPS data, is one of the research issue in robotic community, in last decade. In this thesis, a new and efficient method is proposed for SLAM at the level of objects. The maps produced by state of the art methods don’t have a... 

Visual odometry(VO) is the process of estimating the egomotion of an agent(e.g., vehicle, human, and robot) using the input of a single or multiple cameras attached to it. Application domains include robotics, wearable computing, augmented reality, and automotive. The term was chosen for its similarity to wheel odometry, which incrementally estimates the motion of a vehicle by integrating the number of turns of its wheels over time. Likewise, VO operates by incrementally estimating the pose of the vehicle through examination of the changes that movements induces on the images of its onboard cameras. For the VO to work effectively, there should be sufficient illumination in the environment... 

The highest and most difficult layer of Natural Language Processing, is the understanding of meaning. As a result, lexicons and annotated corpora are of the utmost importance in this area. However, the lack of such semantic resources, especially in Abstract Meaning Representation (AMR), is one of the main issues in this field for Persian Language. This work by modeling PropBank, a semantic valency lexicon for English predicates, is the first step towards building such lexicons for Persian Language with the focus on AMR. Thus, a guideline describing how to annotate the Persian predicates is provided which first evaluates the common structures between the two languages and then focuses on the...