Recent advances of 3D acquisition devices have enabled large-scaleacquisition of 3D scene data. Such data, if completely and well annotated, canserve as useful ingredients for a wide spectrum of computer vision and graphicsworks such as data-driven modeling and scene understanding, object detectionand recognition. However, annotating a vast amount of 3D scene data remainschallenging due to the lack of an effective tool and/or the complexity of 3Dscenes (e.g. clutter, varying illumination conditions). This paper aims tobuild a robust annotation tool that effectively and conveniently enables thesegmentation and annotation of massive 3D data. Our tool works by coupling 2Dand 3D information via an interactive framework, through which users canprovide high-level semantic annotation for objects. We have experimented ourtool and found that a typical indoor scene could be well segmented andannotated in less than 30 minutes by using the tool, as opposed to a few hoursif done manually. Along with the tool, we created a dataset of over a hundred3D scenes associated with complete annotations using our tool. The tool anddataset are available at www.scenenn.net.
展开▼