Scalable active learning for multiclass image classification.

Machine learning techniques for computer vision applications like object recognition, scene classification, etc., require a large number of training samples for satisfactory performance. Especially when classification is to be performed over many categories, providing enough training samples for each category is infeasible. This paper describes new ideas in multiclass active learning to deal with the training bottleneck, making it easier to train large multiclass image classification systems. First, we propose a new interaction modality for training which requires only yes-no type binary feedback instead of a precise category label. The modality is especially powerful in the presence of hundreds of categories. For the proposed modality, we develop a Value-of-Information (VOI) algorithm that chooses informative queries while also considering user annotation cost. Second, we propose an active selection measure that works with many categories and is extremely fast to compute. This measure is employed to perform a fast seed search before computing VOI, resulting in an algorithm that scales linearly with dataset size. Third, we use locality sensitive hashing to provide a very fast approximation to active learning, which gives sublinear time scaling, allowing application to very large datasets. The approximation provides up to two orders of magnitude speedups with little loss in accuracy. Thorough empirical evaluation of classification accuracy, noise sensitivity, imbalanced data, and computational performance on a diverse set of image datasets demonstrates the strengths of the proposed algorithms.

Feasibility of intrafraction whole-body motion tracking for total marrow irradiation.

With image-guided tomotherapy, highly targeted total marrow irradiation (TMI) has become a feasible alternative to conventional total body irradiation. The uncertainties in patient localization and intrafraction motion of the whole body during hour-long TMI treatment may pose a risk to the safety and accuracy of targeted radiation treatment. The feasibility of near-infrared markers and optical tracking system (OTS) is accessed along with a megavoltage scanning system of tomotherapy. Three near-infrared markers placed on the face of a rando phantom are used to evaluate the capability of OTS in measuring changes in the markers' positions as the rando is moved in the translational direction. The OTS is also employed to determine breathing motion related changes in the position of 16 markers placed on the chest surface of human volunteers. The maximum uncertainty in locating marker position with the OTS is 1.5 mm. In the case of normal and deep breathing motion, the maximum marker position change is observed in anterior-posterior direction with the respective values of 4 and 12 mm. The OTS is able to measure surface changes due to breathing motion. The OTS may be optimized to monitor whole body motion during TMI to increase the accuracy of treatment delivery and reduce the radiation dose to the lungs.

Estimating 3D positions and velocities of projectiles from monocular views.

In this paper, we consider the problem of localizing a projectile in 3D based on its apparent motion in a stationary monocular view. A thorough theoretical analysis is developed, from which we establish the minimum conditions for the existence of a unique solution. The theoretical results obtained have important implications for applications involving projectile motion. A robust, nonlinear optimization-based formulation is proposed, and the use of a local optimization method is justified by detailed examination of the local convexity structure of the cost function. The potential of this approach is validated by experimental results.