An electronic device and method for shape refinement of a 3D mesh reconstructed from images is disclosed. A set of images of an object is acquired and used to estimate a first 3D mesh of a head portion of the object. A first set of operations is executed on the first 3D mesh to generate a second 3D mesh. The first set of operations includes a removal of one or more regions which are unneeded for head-shape estimation and/or a removal of one or more mesh artifacts associated with a 3D shape or a topology of the first 3D mesh. A 3D template mesh is processed to determine a set of filling patches which corresponds to a set of holes in the second 3D mesh. Based on the second 3D mesh and the set of filling patches, a hole filling operation is executed to generate a final 3D mesh.

A system for characterising surfaces in a real-world scene, the system comprising an object identification unit operable to identify one or more objects within one or more captured images of the real-world scene, a characteristic identification unit operable to identify one or more characteristics of one or more surfaces of the identified objects, and an information generation unit operable to generate information linking an object and one or more surface characteristics associated with that object.

A system for characterising surfaces in a real-world scene, the system comprising an object identification unit operable to identify one or more objects within one or more captured images of the real-world scene, a characteristic identification unit operable to identify one or more characteristics of one or more surfaces of the identified objects, and an information generation unit operable to generate information linking an object and one or more surface characteristics associated with that object.

A method and apparatus for correcting an error in depth information estimated from a two-dimensional (2D) image are disclosed. The method includes diagnosing an error in depth information by inputting a color image and depth information estimated using the color image to a depth error detection network, and determining enhanced depth information by maintaining or correcting the depth information based on the diagnosed error.

Disclosed is an apparatus for generating a depth map using a monocular image. The apparatus includes: a deep convolution neural network (DCNN) optimized based on an encoder and decoder architecture. The encoder extracts one or more features from the monocular image according to the number of provided feature layers, and the decoder calculates displacements of mismatched pixels from the features extracted from different feature layers, and generates the depth map for the monocular image.

A method of processing an image, an electronic device, and a storage medium. The method includes: determining a shape parameter, a texture parameter and a static wrinkle parameter for an object according to an input image; reconstructing a coarse reconstructed shape for the object by using the shape parameter, and computing a coarse reconstructed texture map for the object by using the texture parameter; determining a fine reconstructed shape and a fine reconstructed texture map according to the static wrinkle parameter, the shape parameter and the texture parameter; and performing a rendering process based on the coarse reconstructed shape, the coarse reconstructed texture map, the fine reconstructed shape and the fine reconstructed texture map, so as to obtain a coarse reconstructed image and a fine reconstructed image for the input image.

A method of processing an image, an electronic device, and a storage medium. The method includes: determining a shape parameter, a texture parameter and a static wrinkle parameter for an object according to an input image; reconstructing a coarse reconstructed shape for the object by using the shape parameter, and computing a coarse reconstructed texture map for the object by using the texture parameter; determining a fine reconstructed shape and a fine reconstructed texture map according to the static wrinkle parameter, the shape parameter and the texture parameter; and performing a rendering process based on the coarse reconstructed shape, the coarse reconstructed texture map, the fine reconstructed shape and the fine reconstructed texture map, so as to obtain a coarse reconstructed image and a fine reconstructed image for the input image.

A depth detection method, a method for training a depth estimation branch network, an electronic device, and a storage medium are provided, which relate to the field of artificial intelligence, particularly to the technical fields of computer vision and deep learning, and may be applied to intelligent robot and automatic driving scenarios. The specific implementation includes: extracting a high-level semantic feature in an image to be detected, wherein the high-level semantic feature is used to represent a target object in the image to be detected; inputting the high-level semantic feature into a pre-trained depth estimation branch network, to obtain distribution probabilities of the target object in respective sub-intervals of a depth prediction interval; and determining a depth value of the target object according to the distribution probabilities of the target object in the respective sub-intervals and depth values represented by the respective sub-intervals.

A model creation apparatus being configured to: hold at least one image of the registration target object in one or more postures and a reference model indicating a shape of a reference object; acquire information indicating a feature of the registration target object in a first posture; and correct, when a shape in the first posture that is indicated by the reference model is determined to be dissimilar based on a predetermined first condition, the reference model based on the information indicating the feature to thereby create the model indicating the shape of the registration target object.

One or more two-dimensional images of a three-dimensional object may be analyzed to estimate a three-dimensional mesh representing the object and a mapping of the two-dimensional images to the three-dimensional mesh. Initially, a correspondence may be determined between the images and a UV representation of a three-dimensional template mesh by training a neural network. Then, the three-dimensional template mesh may be deformed to determine the representation of the object. The process may involve a reprojection loss cycle in which points from the images are mapped onto the UV representation, then onto the three-dimensional template mesh, and then back onto the two-dimensional images.