Figure 1: The skin of a character (a) can be significantly distorted using standard techniques (b), but moves realistically with our method (c). Flexing a hand (d) realistically moves the skin, along with skin properties like normal maps (e). Our method can also be applied to skin-tight clothes (f) and animal skin (g).
AbstractWe present a novel approach for simulating thin hyperelastic skin. Real human skin is only a few millimeters thick. It can stretch and slide over underlying body structures such as muscles, bones, and tendons, revealing rich details of a moving character. Simulating such skin is challenging because it is in close contact with the body and shares its geometry. Despite major advances in simulating elastodynamics of cloth and soft bodies for computer graphics, such methods are difficult to use for simulating thin skin due to the need to deal with non-conforming meshes, collision detection, and contact response. We propose a novel Eulerian representation of skin that avoids all the difficulties of constraining the skin to lie on the body surface by working directly on the surface itself. Skin is modeled as a 2D hyperelastic membrane with arbitrary topology, which makes it easy to cover an entire character or object. Unlike most Eulerian simulations, we do not require a regular grid and can use triangular meshes to model body and skin geometry. The method is easy to implement, and can use low resolution meshes to animate high-resolution details stored in texture-like maps. Skin movement is driven by the animation of body shape prescribed by an artist or by another simulation, and so it can be easily added as a post-processing stage to an existing animation pipeline. We provide several examples simulating human and animal skin, and skin-tight clothes.
Semantic image segmentation is a popular image segmentation technique where each pixel in an image is labeled with an object class. This technique has become a vital part of image analysis nowadays as it facilitates the description, categorization, and visualization of the regions of interest in an image. The recent developments in computer vision algorithms and the increasing availability of large datasets have made semantic image segmentation very popular in the field of computer vision. Motivated by the human visual system which can identify objects in a complex scene very efficiently, researchers are interested in building a model that can semantically segment an image into meaningful object classes. This paper reviews deep learning-based semantic segmentation techniques that use deep neural network architectures for image segmentation of biomedical images. We have provided a discussion on the fundamental concepts related to deep learning methods used in semantic segmentation for the benefit of readers. The standard datasets and existing deep network architectures used in both medical and non-medical fields are discussed with their significance. Finally, this paper concludes by discussing the challenges and future research directions in the field of deep learning-based semantic segmentation for applications in the medical field.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.