Nazlı İkizler

Searching Video for Complex Activities with Finite State Models

We describe a method of representing human activities that allows a collection of motions to be queried without examples, using a simple and effective query language. Our approach is based on units of activity at segments of the body, that can be composed across space and across the body to produce complex queries. The presence of search units is inferred automatically by tracking the body, lifting the tracks to 3D and comparing to models trained using motion capture data. We show results for a large range of queries applied to a collection of complex motion and activity. Our models of short time scale limb behaviour are built using labelled motion capture set. We compare with discriminative methods applied to tracker data; our method offers significantly improved performance. We show experimental evidence that our method is robust to view direction and is unaffected by the changes of clothing.

Searching for Complex Human Activities with No Visual Examples

We describe a method of representing human activities that allows a collection of motions to be queried without examples, using a simple and effective query language. Our approach is based on units of activity at segments of the body, that can be composed across space and across the body to produce complex queries. The presence of search units is inferred automatically by tracking the body, lifting the tracks to 3D and comparing to models trained using motion capture data. Our models of short time scale limb behaviour are built using labelled motion capture set. We show results for a large range of queries applied to a collection of complex motion and activity. We compare with discriminative methods applied to tracker data; our method offers significantly improved performance. We show experimental evidence that our method is robust to view direction and is unaffected by some important changes of clothing.

Recognizing actions from still images

In this paper, we approach the problem of understanding human actions from still images. Our method involves representing the pose with a spatial and orientational histogramming of rectangular regions on a parse probability map. We use LDA to obtain a more compact and discriminative feature representation and binary SVMs for classification. Our results over a new dataset collected for this problem show that by using a rectangle histogramming approach, we can discriminate actions to a great extent. We also show how we can use this approach in an unsupervised setting. To our best knowledge, this is one of the first studies that try to recognize actions within still images.

Human action recognition with line and flow histograms

We present a compact representation for human action recognition in videos using line and optical flow histograms. We introduce a new shape descriptor based on the distribution of lines which are fitted to boundaries of human figures. By using an entropy-based approach, we apply feature selection to densify our feature representation, thus, minimizing classification time without degrading accuracy. We also use a compact representation of optical flow for motion information. Using line and flow histograms together with global velocity information, we show that high-accuracy action recognition is possible, even in challenging recording conditions.

Re-ranking of image search results using a graph algorithm

Although one of the most common usages of Internet is searching, especially in image search, the users are not satisfied due to many irrelevant results. In this paper we present a method to identify irrelevant results of image search on the Internet and re-rank the results so that the relevant results will have a higher priority within the list. The proposed method represents the similarity of images in a graph structure, and then finds the densest component in the graph representing the most similar set of images corresponding to the query.

Person search made easy

In this study, we present a method to extensively reduce the number of retrieved images and increase the retrieval performance for the person queries on the broadcast news videos. A multi-modal approach which integrates face and text information is proposed. A state-of-the-art face detection algorithm is improved using a skin color based method to eliminate the false alarms. This pruned set is clustered to group the similar faces and representative faces are selected from each cluster to be provided to the user. For six person queries of TRECVID2004, on the average, the retrieval rate is increased from 8% to around 50%, and the number of images that the user has to inspect are reduced from hundreds and thousands to tens.

Diagnosis of gastric carcinoma by classification on feature projections

A new classification algorithm, called benefit maximizing classifier on feature projections (BCFP), is developed and applied to the problem of diagnosis of gastric carcinoma. The domain contains records of patients with known diagnosis through gastroscopy results. Given a training set of such records, the BCFP classifier learns how to differentiate a new case in the domain. BCFP represents a concept in the form of feature projections on each feature dimension separately. Classification in the BCFP algorithm is based on a voting among the individual predictions made on each feature. In the gastric carcinoma domain, a lesion can be an indicator of one of nine different levels of gastric carcinoma, from early to late stages. The benefit of correct classification of early levels is much more than that of late cases. Also, the costs of wrong classifications are not symmetric. In the training phase, the BCFP algorithm learns classification rules that maximize the benefit of classification. In the querying phase, using these rules, the BCFP algorithm tries to make a prediction maximizing the benefit. A genetic algorithm is applied to select the relevant features. The performance of the BCFP algorithm is evaluated in terms of accuracy and running time. The rules induced are verified by experts of the domain.

Maximizing Benefit of Classifications Using Feature Intervals

There is a great need for classification methods that can properly handle asymmetric cost and benefit constraints of classifications. In this study, we aim to emphasize the importance of classification benefits by means of a new classification algorithm, Benefit-Maximizing classifier with Feature Intervals (BMFI) that uses feature projection based knowledge representation. Empirical results show that BMFI has promising performance compared to recent cost-sensitive algorithms in terms of the benefit gained.

Improvement of face detection algorithms for news videos

People are the most important subjects in news videos and for proper retrieval of person images, face detection is a very crucial step. However, face detection and recognition in news videos is a very challenging task due to the huge irregularities and high noise level in the data. This study presents a method that combines skin detection and Schneiderman-Kanade face detection, for improving the face detection performance in news videos for a better retrieval. This method has been tested on TRECVID 2003 dataset and the results are very promising.