Photo-editing software restricts the control of objects in a photograph to the 2D image plane. We present a method that enables users to perform the full range of 3D manipulations, including scaling, rotation, translation, and nonrigid deformations, to an object in a photograph. As 3D manipulations often reveal parts of the object that are hidden in the original photograph, our approach uses publicly available 3D models to guide the completion of the geometry and appearance of the revealed areas of the object. The completion process leverages the structure and symmetry in the stock 3D model to factor out the effects of illumination, and to complete the appearance of the object. We demonstrate our system by producing object manipulations that would be impossible in traditional 2D photo-editing programs, such as turning a car over, making a paper-crane flap its wings, or manipulating airplanes in a historical photograph to change its story.

No abstract

Abstract. In this paper, a four-dimensional spatiotemporal shape context descriptor is introduced and used for human activity recognition in video. The spatiotemporal shape context is computed on silhouette points by binning the magnitude and direction of motion at every point with respect to given vertex, in addition to the binning of radial displacement and angular offset associated with the standard 2D shape context. Human activity recognition at each video frame is performed by matching the spatiotemporal shape context to a library of known activities via k-nearest neighbor classification. Activity recognition in a video sequence is based on majority classification of the video frame results. Experiments on the Weizmann set of ten activities indicate that the proposed shape context achieves better recognition of activities than the original 2D shape context, with overall recognition rates of 90% obtained for individual frames and 97.9% for video sequences.

No abstract