Nowadays, huge amounts of visual data, e.g., videos and images, have become widely accessible. Therefore, intelligently categorizing the large and growing collections of data for access convenience has been a central goal for modern computer vision research.
As a result of this flow and an increased security awareness for access control systems in parking lots, buildings and restricted areas, the need for vehicle identification and classification has become prevalent. In this seminar, we present a framework of vehicle make and model recognition (VMMR) that uses Multiple Instance Learning (MIL) for fine-grained classification.
In MIL, we are usually confronted with a large instance space for even moderately sized datasets since each bag may contain many instances. Hence it is important to design efficient instance pruning techniques to speed up the learning process without compromising on the performance. To meet these objectives, we developed a multiple-instance learning scheme using a selection strategy to identify the most representative examples in the positive and negative training bag.
Our solution lies in constraining the high dimensional instance space by selecting a more concise set of visual patches instead of the traditional coarse-grained instance labeling. Our method is generic in the sense that it can be used with any local features or feature sets representing the content of an image region.
Faezeh Tafazzoli is a doctoral candidate in the Multimedia Research Lab, in the department of Computer Science and Computer Engineering (CECS), University of Louisville. She received her Masters degrees from University of Nevada, Reno and Amirkabir University of Technology, Iran. Her research is focused on the fields of Computer Vision and Machine Learning. In particular, she is interested in fine-grained classification, object detection and recognition, content-based image retrieval, and deep vision. Faezeh has worked as a research scientist intern at Xerox Innovation Group (XRCW), PARC and Eye-Com Corporation in the areas of human gait analysis, visual remote assessment of Parkinson's disease and gaze tracking.