60 results about "3d geometry" patented technology

The invention discloses a three-dimensional reconstruction method based on a pulse camera and electronic equipment, and relates to the technical field of three-dimensional reconstruction. According to the method, the problems of insufficient spatial-temporal information modeling and inaccurate initial attitude estimation in pulse camera three-dimensional reconstruction in related technologies are effectively solved, the precision of a three-dimensional reconstruction model is improved, a three-dimensional scene with higher quality and more realistic sense can be generated, meanwhile, efficient alignment of pulse data and three-dimensional geometry is realized through joint optimization of a framework, and the accuracy of the three-dimensional reconstruction model is improved. And the overall robustness and generalization ability of the system are enhanced.

A method of implementing directional opacity handling in Gaussian splats is disclosed. An approximate representation of a scene is received. 3D geometry data is generated. The 3D geometry data includes an approximate position and normal of each splat in a set of splats in the scene. Directional opacity values are calculated for the set of splats based on an application of a machine-learning model trained to optimize values of splat parameters, the values including the directional opacity values. The calculated directional opacity values are applied to the set of splats to enhance processing of the set of splats.

The present invention relates to the technical field of buildings, and in particular, to a visual inspection and measurement method for a reinforcing mesh, comprising the steps of: using an RGBD camera to perform data acquisition on a building site; performing Mask R-CNN instance segmentation on RGB, preliminarily identifying a workpiece, and segmenting a 2D workpiece from the background; converting the obtained 2D workpiece into a 3D point cloud on the basis of an intrinsic mapping relationship of the RGBD camera; and for on-site acquired data, obtaining a segmentation result of rebar details by means of FPFH feature matching with an offline model; and on the basis of an identified rebar point cloud, performing cylinder fitting to obtain the number of rebars within the field of view. According to the present invention, preliminary identification of rebars is performed on a 2D site image by means of Mask R-CNN instance segmentation, and then rebar 3D geometry is precisely identified by using FPFH feature consistency.

A technique for generating a sequence of geometries includes converting, via an encoder neural network, one or more input geometries corresponding to one or more frames within an animation into one or more latent vectors. The technique also includes generating the sequence of geometries corresponding to a sequence of frames within the animation based on the one or more latent vectors. The technique further includes causing output related to the animation to be generated based on the sequence of geometries.

Optical sensors (e.g., cameras) and (e.g., IR) illumination sources may be distributed in an environment (e.g., an interior space such as a cabin or cockpit of an ego-machine) and synchronized to generate frames of sensor data, which may be used to reconstruct 3D geometry and / or 3D pose of an occupant, operator, or other object in the environment. For example, stereo vision may be used to generate one or more depth maps from image data generated using different cameras, the depth map(s) may be transformed into a 3D point cloud, and surface reconstruction may be applied to reconstruct the 3D geometry of surface(s) in the environment. A 3D pose, one or more keypoints (e.g., facial landmarks), or some other representation of the shape of the reconstructed surface(s) may be extracted from the reconstructed surface and used in one or more downstream tasks, such as driver and / or occupant monitoring tasks.

One embodiment of the present invention sets forth a technique for training a machine learning model on a geometry generation task. The technique includes generating, via execution of a diffusion model, a first set of training output corresponding to a first set of three-dimensional (3D) geometries based on a first set of conditioning inputs associated with a first conditioning mode, and training the diffusion model based on a first set of loss values associated with the first set of training output. The technique further includes generating, via execution of the diffusion model and a first adapter model, a second set of training output corresponding to a second set of 3D geometries based on a second set of conditioning inputs associated with a second conditioning mode, and training the first adapter model based on a second set of loss values associated with the second set of training output.

The invention discloses a visual angle controllable continuous image generation system and method based on three-dimensional point clouds, and relates to the technical field of crossing of computer vision and graphics, the visual angle controllable continuous image generation system comprises a 3D geometric coding module, a condition rendering module and a video generation module, the 3D geometric coding module is used for constructing and updating dynamic 3D scene representation in real time, and the condition rendering module is used for rendering the dynamic 3D scene representation in real time; the condition rendering module is used for generating 2D rendering frames with geometric constraints, the video generation module completes video synthesis based on a diffusion model, and for continuous frame requirements, feature matching information is injected and cache is updated in a blocking manner by combining a ControlNet subnet. By means of a bidirectional projection mechanism, on the basis that 3D point cloud is rendered into 2D frames with holes through forward projection, 3D cache of the complemented RGB-D frames is reversely updated through reverse projection, dynamic balance of geometric consistency and content updating is ingeniously achieved, the content can be updated in time while the geometric structure of the generated image is kept accurate, and the image quality is improved. And the authenticity and the continuity of the image are enhanced.

In various examples, optical depth estimation for interior space monitoring systems and applications is disclosed. Absolute 3D depth estimates from monocular image data may be generated using a machine learning model using 3D geometry priors and a joint learning framework that combines depth estimation with object segmentation. Three-dimensional geometry priors provide surface-level information that enriches the model's understanding of the relevant spatial geometry and resolves scale ambiguity in monocular depth estimation within automotive in-cabin environments. The model may include a common (e.g., shared) encoder stage that outputs features extracted from an optical image sensor feed to separate decoder stages that include a depth estimation decoder and a segmentation decoder. Joint learning for depth estimation and segmentation tasks during training achieves a more nuanced understanding of the in-cabin environment, leading to significantly improved depth accuracy.

Optical sensors (e.g., cameras) and (e.g., IR) illumination sources may be distributed in an environment (e.g., an interior space such as a cabin or cockpit of an ego-machine) and synchronized to generate frames of sensor data, which may be used to reconstruct 3D geometry and / or 3D pose of an occupant, operator, or other object in the environment. For example, stereo vision may be used to generate one or more depth maps from image data generated using different cameras, the depth map(s) may be transformed into a 3D point cloud, and surface reconstruction may be applied to reconstruct the 3D geometry of surface(s) in the environment. A 3D pose, one or more keypoints (e.g., facial landmarks), or some other representation of the shape of the reconstructed surface(s) may be extracted from the reconstructed surface and used in one or more downstream tasks, such as driver and / or occupant monitoring tasks.

The invention discloses an image adjustment 3D content creation architecture based on 2D diffusion, and relates to the technical field of diffusion models, and the architecture comprises a diffusion model which is composed of a plurality of cross attention blocks in a U-Net structure and supports effective fusion of various modes of texts, images and camera parameters. In the present invention, it is devoted to create 3D content using a potential diffusion model. The 3D geometry is not generated directly through a potential diffusion model, but a two-stage formula is employed. First, a potential diffusion model of view conditioned reflex is organized, and a multi-view image is synthesized with a monocular image and the outside of a camera as inputs. Next, a neural radiation field is trained using the synthesized multi-view image, which is easily optimized as volume rendering is differentiable. After the neural radiation field training is completed, a 3D geometric model is generated through an advancing cube algorithm applied to a density field. The framework eliminates the requirement for pairing 3D training data and does not require a large amount of computing resources.

An electronic apparatus for training an AI (Artificial Intelligence) model to generate an image of an object based on an input pose comprises at least one processor including processing circuitry and memory storing instructions. The instructions, when executed by the at least one processor individually or collectively, cause the electronic apparatus to obtain a plurality of images for training the AI model, obtain training images including an object from the plurality of images, obtain at least one parameter corresponding to each of the training images, obtain 3D geometry information of the object based on the at least one parameter, obtain a spherical neural texture mapped with the 3D geometry information, obtain warping prediction information for the spherical neural texture, update the spherical neural texture based on the warping prediction information, and train the AI model based on the updated spherical neural texture.

A method, system, and apparatus for computer-aided design and manufacture of a physical structure using hybrid additive and subtractive manufacturing, including media-encoded computer program products, includes, in one aspect, a method comprising: obtaining data on the 3D geometry of a part; simulating at least a portion of a manufacturing process, said at least a portion including adding a first material in a first stage and removing a second material in a second subsequent stage, wherein the second material comprises a portion of the first material, removing the second material includes mixing between the materials added in the first and second stages, and simulating the thermal effects of adding and removing materials in the first and second stages; and adjusting the amount of said portion based on the simulation results to prevent the part from deviating from its three-dimensional geometry, said deviation resulting in insufficient material available for mixing.

A computer-implemented method for marking an object (5) on a 3D image, comprising the steps of: providing a 3D image comprising an object (5); displaying at least one 2D slice (10) of the 3D image to a user and locating a 3D geometry within the 3D image; receiving location information of the 3D geometry selected by the user; automatically defining a 3D region of interest (7) within the 3D image comprising the 3D geometry and cropping image data from the 3D region of interest (7) from the 3D image; a trained algorithm is applied to the image data from the 3D region of interest (7), where the output of the trained algorithm is a 3D volume (9) within the 3D region of interest (7) predicted to cover the object (5).

Generating a rendered image comprises: generating 904 a segmentation mask for an input 3D geometry; generating a plurality of texture maps associated with portions of the geometry; generating 906 a pl

In an example method, an NWM molding blank comprising a nonwoven material held compressed by an adhesive is placed in a separable mold and heated to a melting temperature of the adhesive. The molding blank expands, shaping into an NWM molded object having a 3D geometry as an intermediate NWM object. The intermediate NWM object is cooled through a temperature band having an upper boundary and a lower boundary, and further cooled to a solidification temperature of the adhesive. The upper boundary is above the solidification temperature and the lower boundary is below the adhesive melting temperature. While in the temperature band, the mold is separated so that exposed surfaces of the intermediate NWM object are accessible. The object is then transported to a contoured forming surface of a forming bed by an actuatable arm having an end effector that grips the exposed surfaces via vacuum suction, the object is lifted from the mold and placed on the contoured forming surface. Optionally, the end effector contacting surfaces include final forming features. The actuatable arm compresses the intermediate NWM object against the contoured forming surface and, optionally, the end effector final forming features, and continues compression until cooled to the solidification temperature.

The present invention concerns a system and a method for collision avoidance of a 3D robotic concrete printer, whereby the method comprises the steps of performing an application and / or manipulation process by executing instructions, such as based on G-Code, with a 3D robotic concrete printer by moving a tool in a path for applying and / or manipulating concrete material; and moving the tool in response to commands to a motion planner, said movement being assisted by a model-based collision prediction system, whereby the model-based collision-prediction system involves the steps of inferring and generating 3D geometries of concrete structures from applying and / or manipulating concrete structures by commanding the 3D robotic concrete printer, using the 3D geometries as a 3D collision model, which allow at any point in time to check if parts of the 3D robotic concrete printer would yield a collision when performing said movement; whereby the 3D collision model of the concrete structure is developed in parallel to the application and / or manipulation process and is updated in correspondence to the progress of the execution of the process instructions.

Generating a rendered image comprises: generating 904 a segmentation mask for an input 3D geometry; generating a plurality of texture maps associated with portions of the geometry; generating 906 a pl

Unsupervised volumetric 3D animation (UVA) of non-rigid deformable objects without annotations learns the 3D structure and dynamics of objects solely from single-view red / green / blue (RGB) videos and decomposes the single-view RGB videos into semantically meaningful parts that can be tracked and animated. Using a 3D autodecoder framework, paired with a keypoint estimator via a differentiable perspective-n-point (PnP) algorithm, the UVA model learns the underlying object 3D geometry and parts decomposition in an entirely unsupervised manner from still or video images. This allows the UVA model to perform 3D segmentation, 3D keypoint estimation, novel view synthesis, and animation. The UVA model can obtain animatable 3D objects from a single or a few images. The UVA method also features a space in which all objects are represented in their canonical, animation-ready form. Applications include the creation of lenses from images or videos for social media applications.

A system and method for viewing aerial imagery and derived content on a web-based imagery browser that can be accessed by at least two users simultaneously. In addition, a system for interactively generating 3D geometry data for a location imaged from multiple viewpoints.

Systems and methods for 3D image scanners for real-time dynamic 3D surface imaging are disclosed. Embodiments of the present systems and methods describe a system and method comprising a first and / or second camera, and a projector, and a processor, the projector projects structured light with a fringe pattern onto a 3D object, the processor is configured to extract a phase map and a texture image from the image, and the processor calculates depth information from the phase map. Embodiments further describe methods and systems to determine wrapped phase from the image using a Hilbert transform, and to generate absolute phase from the wrapped phase using a combination of a quality-guided path following algorithm, a two-wavelength phase unwrapping algorithm, or a Markov Random Field method, and to generate a phase map from the absolute phase to determine depth information of the 3D object. Algorithms using conformal mapping, optimal transport mapping, and Tschupmueller mapping are used to register and track captured 3D geometry surfaces.

Systems, methods, and apparatuses for controlling a robot including a manipulator, including: determining three-dimensional (3D) geometry information about a target object based on an image of the target object; determining 3D geometry information about a scene in which the target object is to be placed based on at least one image of the scene; obtaining affordance information by providing the 3D geometry information about the target object and the 3D geometry information about the scene to at least one neural network model; commanding the robot to grasp the target object using the manipulator according to a grasp orientation corresponding to the affordance information; and commanding the robot to position the manipulator according to a placement direction corresponding to the affordance information in order to place the target object at a location in the scene.

The present disclosure relates to object relationship estimation from 3D semantic meshes. The implementations disclosed herein provide systems and methods for determining relationships between objects based on a raw semantic mesh representing vertices and faces of a 3D geometry of a physical environment. Such raw semantic meshes can be generated and used to provide input to a machine learning model that estimates relationships between objects in the physical environment. For example, the machine learning model can output a graph of nodes and edges that indicates that a vase is on a table or that a particular instance V1 of a vase is on a particular instance T1 of a table.

The present invention provides a method for aligning a 2D image and a floor plan based on 3D geometry, comprising: (a) acquiring a 2D image taken through a camera installed in a specific space; and an image alignment device aligning the 2D image and the floor plan based on a neural network geometry estimation model. (b) generating an image 3D coordinate system by extending the 2D image into a 3D space with reference to 3D coordinate values ​​for each pixel of the 2D image inferred through a 3D model, and generating a floor plan 3D coordinate system by extending the 2D image into 3D space with a plane of the floor plan for the specific space as a floor plane; (b) generating alignment information between the image 3D coordinate system and the floor plan 3D coordinate system with reference to plane correspondence information between a first plane recognized in the image 3D coordinate system and a second plane in the floor plan 3D coordinate system corresponding to the first plane, and line correspondence information between 3D points in the image 3D coordinate system and 2D points in the floor plan 3D coordinate system corresponding to the 3D points; and (c) performing alignment between the 2D image and the floor plan by projecting each of the pixels of the 2D image onto the floor plan with reference to the alignment information.

The invention concerns a method (1) to detect deformations of, and / or damages to, structures permanently arranged on the earth's surface. In particular, said method (1) comprises: acquiring (11) georeferencing data indicative of geographical reference positions of predefined points of interest of a given structure to be monitored permanently arranged on the earth's surface, wherein said predefined points of interest are representative of a 3D geometry of the given structure without deformations and damages; acquiring (12) a SAR image of an area of the earth's surface where the given structure is arranged, wherein said SAR image is associated with a given reference coordinate system; transforming (13) the geographical reference positions of the predefined points of interest into corresponding expected positions in the given reference coordinate system associated with the acquired SAR image so as to carry out a reprojection of the 3D geometry of the given structure without deformations and damages on the acquired SAR image; identifying (14) in the acquired SAR image the predefined points of interest of the given structure; determining (15) actual positions in the given reference coordinate system associated with the acquired SAR image of the predefined points of interest identified in said SAR image; making a comparison (16) between the expected positions of the predefined points of interest and the corresponding actual positions in the acquired SAR image; and detecting (17) one or more deformations of, and / or one or more damages to, said given structure on the basis of the comparison made.

An automated cage-to-cage fitting technique is used to fit arbitrary three-dimensional (3D) geometry to arbitrary target 3D geometry for 3D avatars in a 3D environment, such as an online game. Each graphical representation of items of clothing is provided with an inner cage and with an outer cage. The inner cage of a clothing item (a current layer) is mapped to the outer cage of another clothing item (a previous / underlying layer) that has already been fitted on the 3D avatar. This mapping enables the current layer to conform to the shape of the previous layer. The technique allows any body geometry to be fitted with any clothing geometry, including enabling layers of clothing to be fitted over underlying layer(s) of clothing, thereby providing customization.

The present disclosure provides an electrophysiology system to facilitate visualizing a proximity of at least one catheter electrode to a 3D geometry of a biological tissue. The system includes a computing device including at least one processor in communication with a memory, wherein the processor is configured to determine the proximity between the at least one catheter electrode and the biological tissue using at least one measurement. The system further includes a display device configured to display the 3D geometry of the biological tissue and a visual effect illustrating the proximity between the at least one catheter electrode and the biological tissue.

Rendering 3D geometry involves processing a very large amount of geometry. Compression techniques can be used to decrease the amount of data required for such geometry overall. A particular compression format for geometry is dense compression format, in which triangle strips are represented in highly compacted code sequences. In particular, compression code sequences describe the connectivity between triangles of a strip, thus ultimately provide a compact representation of which vertex indices comprise each triangle. A vertex index is an index into a vertex buffer that stores the actual vertex data, allowing for deduplication of such data. Though compact, such compression code sequences are somewhat tricky to decompress. A technique is provided herein for decompressing such code sequences. In particular, the technique involves a series of bitwise, arithmetic, and / or logical operations that expand out the code sequences into indices for the triangles.

A method is provided for using Light-Field Dynamic Homography (DH) to generate a large virtual focal plane from a non-rigid camera array with narrow overlaps between their fields of view (FOV). The method incorporates the 3D geometry of the cameras and employs non-linear least square optimization to dynamically estimate the inter-view homography transformations. Remarkably, only two feature correspondences are required between adjacent views to stitch the images and generate a wide virtual focal plane array, eliminating the need for significant FOV overlaps between the multiple cameras.

The invention relates to a method for generating an extended FEM model (570) with a 3D geometry of a formed technical component (10), in particular a sheet metal component for a vehicle body, for an FEM simulation for calculating, testing and / or validating design parameters of the technical component (10), wherein the extended FEM model (570) comprises a forming profile (380) of at least one forming parameter (370) occurring during a forming of the technical component (10), comprising the following steps: - Input of at least one FEM model (270); - Performing a modeling of a spatial forming profile (380) of at least one forming parameter (370) by a trained AI model (350), - Generating an extended FEM model (570) with the 3D geometry of the component (10) and the integrated spatial forming profile (380); - Performing an FEM simulation of the component (10) for the calculation, testing and / or validation of design parameters of the component (10), in particular a crash simulation, using the extended FEM model (570).

A method includes obtaining scan data of a dental object at a local device, wherein the scan data includes primary scan data, wherein the scan data includes secondary scan data indicative of characteristics other than the 3D geometry of the dental object. The method includes establishing a connection between the local device and a remote computing device. The method includes obtaining a performance indicator by detecting a performance of the connection. The method includes, if the performance indicator exceeds a primary performance threshold, entering a full connection mode. The method includes, if the performance indicator is less than the primary performance threshold entering a partial connection mode.