Abstract: A method of generating a training data set for training an image matting machine learning model includes receiving a plurality of foreground images, generating a plurality of composited foreground images by compositing randomly selected foreground images from the plurality of foreground images, and generating a plurality of training images by compositing each composited foreground image with a randomly selected background image. The training data set includes the plurality of training images.

Abstract: In various embodiments, a relevance application quantifies how a trained neural network operates. In operation, the relevance application generates a set of input distributions based on a set of input points associated with the trained neural network. Each input distribution is characterized by a mean and a variance associated with a different neuron included in the trained neural network. The relevance application propagates the set of input distributions through a probabilistic neural network to generate at least a first output distribution. The probabilistic neural network is derived from at least a portion of the trained neural network. Based on the first output distribution, the relevance application computes a contribution of a first input point included in the set of input points to a difference between a first output point associated with a first output of the trained neural network and an estimated mean prediction associated with the first output.

Abstract: In various embodiments, a differentiable rendering application enables an inverse rendering application to infer attributes associated with a 3D scene. In operation, the differentiable rendering application renders an image based on a first set of points associated with the 3D scene. The differentiable rendering application then generates an artificial gradient that approximates a change in a value of a first pixel included in the image with respect to a change in an attribute of a first point included in the first set of points. Subsequently, the inverse rendering application performs optimization operation(s) on the first point based on the artificial gradient to generate a second set of points. Notably, an error associated with the second set of points is less than an error associated with the first set of points.

Abstract: A method of generating a training data set for training an image matting machine learning model includes receiving a plurality of foreground images, generating a plurality of com posited foreground images by com positing randomly selected foreground images from the plurality of foreground images, and generating a plurality of training images by compositing each composited foreground image with a randomly selected background image. The training data set includes the plurality of training images.

Abstract: Techniques are disclosed for image matting. In particular, embodiments decompose the matting problem of estimating foreground opacity into the targeted subproblems of estimating a background using a first trained neural network, estimating a foreground using a second neural network and the estimated background as one of the inputs into the second neural network, and estimating an alpha matte using a third neural network and the estimated background and foreground as two of the inputs into the third neural network. Such a decomposition is in contrast to traditional sampling-based matting approaches that estimated foreground and background color pairs together directly for each pixel. By decomposing the matting problem into subproblems that are easier for a neural network to learn compared to traditional data-driven techniques for image matting, embodiments disclosed herein can produce better opacity estimates than such data-driven techniques as well as sampling-based and affinity-based matting approaches.

Abstract: According to one implementation, a pose estimation and body tracking system includes a computing platform having a hardware processor and a system memory storing a software code including a tracking module trained to track motions. The software code receives a series of images of motion by a subject, and for each image, uses the tracking module to determine locations corresponding respectively to two-dimensional (2D) skeletal landmarks of the subject based on constraints imposed by features of a hierarchical skeleton model intersecting at each 2D skeletal landmark. The software code further uses the tracking module to infer joint angles of the subject based on the locations and determine a three-dimensional (3D) pose of the subject based on the locations and the joint angles, resulting in a series of 3D poses. The software code outputs a tracking image corresponding to the motion by the subject based on the series of 3D poses.

Abstract: In various embodiments, a differentiable rendering application enables an inverse rendering application to infer attributes associated with a 3D scene. In operation, the differentiable rendering application renders an image based on a first set of points associated with the 3D scene. The differentiable rendering application then generates an artificial gradient that approximates a change in a value of a first pixel included in the image with respect to a change in an attribute of a first point included in the first set of points. Subsequently, the inverse rendering application performs optimization operation(s) on the first point based on the artificial gradient to generate a second set of points. Notably, an error associated with the second set of points is less than an error associated with the first set of points.

Abstract: Techniques are disclosed for estimating poses from images. In one embodiment, a machine learning model, referred to herein as the “detector,” is trained to estimate animal poses from images in a bottom-up fashion. In particular, the detector may be trained using rendered images depicting animal body parts scattered over realistic backgrounds, as opposed to renderings of full animal bodies. In order to make appearances of the rendered body parts more realistic so that the detector can be trained to estimate poses from images of real animals, the body parts may be rendered using textures that are determined from a translation of rendered images of the animal into corresponding images with more realistic textures via adversarial learning. Three-dimensional poses may also be inferred from estimated joint locations using, e.g., inverse kinematics.

Abstract: Techniques are disclosed for image matting. In particular, embodiments decompose the matting problem of estimating foreground opacity into the targeted subproblems of estimating a background using a first trained neural network, estimating a foreground using a second neural network and the estimated background as one of the inputs into the second neural network, and estimating an alpha matte using a third neural network and the estimated background and foreground as two of the inputs into the third neural network. Such a decomposition is in contrast to traditional sampling-based matting approaches that estimated foreground and background color pairs together directly for each pixel. By decomposing the matting problem into subproblems that are easier for a neural network to learn compared to traditional data-driven techniques for image matting, embodiments disclosed herein can produce better opacity estimates than such data-driven techniques as well as sampling-based and affinity-based matting approaches.

Abstract: According to one implementation, a pose estimation and body tracking system includes a computing platform having a hardware processor and a system memory storing a software code including a tracking module trained to track motions. The software code receives a series of images of motion by a subject, and for each image, uses the tracking module to determine locations corresponding respectively to two-dimensional (2D) skeletal landmarks of the subject based on constraints imposed by features of a hierarchical skeleton model intersecting at each 2D skeletal landmark. The software code further uses the tracking module to infer joint angles of the subject based on the locations and determine a three-dimensional (3D) pose of the subject based on the locations and the joint angles, resulting in a series of 3D poses. The software code outputs a tracking image corresponding to the motion by the subject based on the series of 3D poses.

Abstract: Techniques are disclosed for estimating poses from images. In one embodiment, a machine learning model, referred to herein as the “detector,” is trained to estimate animal poses from images in a bottom-up fashion. In particular, the detector may be trained using rendered images depicting animal body parts scattered over realistic backgrounds, as opposed to renderings of full animal bodies. In order to make appearances of the rendered body parts more realistic so that the detector can be trained to estimate poses from images of real animals, the body parts may be rendered using textures that are determined from a translation of rendered images of the animal into corresponding images with more realistic textures via adversarial learning. Three-dimensional poses may also be inferred from estimated joint locations using, e.g., inverse kinematics.

Abstract: An image processor inputs a first image and outputs a downscaled second image by upscaling the second image to a third image, wherein the third image is substantially the same size as the first image size with a third resolution, associating pixels in the second image with a corresponding group of pixels from the third set of pixels, sampling a first image area at a first location of the first set of pixels to generate a first image sample, sampling a second image area of the third set of pixels to generate a second image sample, measuring similarity between the image areas, generating a perceptual image value, recursively adjusting values of third set of pixels until the image perception value matches a perceptual standard value, and adjusting pixel values in the second image to a representative pixel value of each of the corresponding group of pixels.

Abstract: An image processor inputs a first image and outputs a downscaled second image by upscaling the second image to a third image, wherein the third image is substantially the same size as the first image size with a third resolution, associating pixels in the second image with a corresponding group of pixels from the third set of pixels, sampling a first image area at a first location of the first set of pixels to generate a first image sample, sampling a second image area of the third set of pixels to generate a second image sample, measuring similarity between the image areas, generating a perceptual image value, recursively adjusting values of third set of pixels until the image perception value matches a perceptual standard value, and adjusting pixel values in the second image to a representative pixel value of each of the corresponding group of pixels.