Abstract: We propose an algorithm for semantic segmentation based on 3D point clouds derived from ego-motion. We motivate five simple cues designed to model specific patterns of motion and 3D world structure that vary with object category. We introduce features that project the 3D cues back to the 2D image plane while modeling spatial layout and context. A randomized decision forest combines many such features to achieve a coherent 2D segmentation and recognize the objec t categories present. Our main contribution is to show how semantic segmentation is possible based solely on motion-derived 3D world structure. Our method works well on sparse, noisy point clouds, and unlike existing approaches, does not need appearance-based descriptors. Experiments were performed on a challenging new video database containing sequences filmed from a moving car in daylight and at dusk. The results confirm t hat indeed, accurate segmentation and recognition are possible using only motion and 3D world structure. Further, we show that the motion-derived informati on complements an existing state-of-the-art appearance-based method, improving both qualitative and quantitative performance.

Abstract: This video presents an immersive environment that allows collaborative and remote 3D interactions. The environment uses multiple camera systems and 3D modeling tools to build, in real time, virtual models of users. Such users can be on distant sites and still share the same virtual environment. Their 3D models are embedded into the virtual environment where they can interact with shared virtual objects. Models encode geometric information that are plugged into a physical simulation for interactive purposes. They also encode photometric information that are used through textures to ensure a good sense of presence.

Abstract: We demonstrate the use of a new representation of video for facilitating a number of common editing tasks. The representation has some of the power of a full reconstruction of 3D surface models from video, but is designed to be easy to recover from a priori unseen and uncalibrated footage. By modelling the im age-formation process as a 2D-to-2D transformation from an object's texture map to the image, modulated by an object-space occlusion mask, we can recover a representation which we term the unwrap mosaic. Many editing operations can be performed on the unwrap mosaic, and then re-composited into the original sequence, for example resizing objects, repainting textures, copying/cutting/pasting objects, and attaching effects layers to deforming objects.

Abstract: Demo sessions of computer vision systems are ideal venues to disseminate scientific results. Existing demo sessions, however, fail to engage the audience f ully. We argue that simultaneous and real-time evaluation of several computer vision systems in a single showcase increases impact. This movie gives an imp ression of such a showcase, namely the VideOlympics, a competition for video search engines. The major aim of the VideOlympics is promoting research in vid eo retrieval. An additional main goal of the VideOlympics is giving the audience a good perspective on the possibilities and limitations of current state-o f-the-art visual search engines. Where traditional evaluation campaigns like the Pascal Visual Object Classes Challenge and TRECVID focus primarily on the effectiveness of collected retrieval results, the VideOlympics also allows taking into account the influence of interaction mechanisms and the advanced vis ualizations in the interface. In the end, all scientists go home with a golden retriever award, but the real winner is the audience.