Neural network processing for multi-object 3D modeling
Embodiments are directed to neural network processing for multi-object three-dimensional (3D) modeling. An embodiment of a computer-readable storage medium includes executable computer program instructions for obtaining data from multiple cameras, the data including multiple images, and generating a 3D model for 3D imaging based at least in part on the data from the cameras, wherein generating the 3D model includes one or more of performing processing with a first neural network to determine temporal direction based at least in part on motion of one or more objects identified in an image of the multiple images or performing processing with a second neural network to determine semantic content information for an image of the multiple images.
1. A non-transitory computer-readable storage medium having stored thereon executable computer program instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:
generating a three-dimensional (3D) object model of each of multiple objects in a scene for production of 3D imaging, including:
obtaining image data from a plurality of cameras, the data comprising a plurality of images of the multiple objects captured by the plurality of cameras;
performing processing of the image data to identify and separate each of the multiple objects, including at least the following:
processing the image data with a first neural network to generate semantic content information for one or more of the plurality of images, wherein the generated semantic content information includes identification of one or more types of objects represented by the multiple objects, one or more types of activities that are occurring in the image associated with the multiple objects, or both, and
processing the image data with a second neural network to identify a background and the multiple objects in an image of plurality of images, to determine a temporal direction for each of the multiple objects based at least in part on motion of the multiple objects identified in the image, the temporal direction describing a 3D motion of each of the multiple objects in the image at a point in time, and to generate a separate model for the background and for each of the multiple objects, the model of the background being static and the model of each of the multiple objects including the respective temporal direction for the object; and
generating the 3D object models for the multiple objects based at least in part on the image data and the generated information regarding the identified one or more types of objects, one or more types of activities, or both and on the determined temporal direction of each of the one or more objects.
2. The medium of claim 1 , further comprising executable computer program instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:
refining a 3D image utilizing the separate models for the background and each of the multiple objects.
3. A system comprising:
one or more processor cores;
a memory to store data for three-dimensional (3D) imaging, the data comprising images; and
inputs from a plurality of cameras for 3D data capture;
wherein the system is to generate a 3D object model of each of multiple objects in a scene for production of 3D imaging, including at least the following:
receiving image data including a plurality of images from the plurality of cameras;
performing processing of the image data to identify and separate each of the multiple objects, including at least the following:
processing the image data with a first neural network to generate semantic content information for one or more of the plurality of images, wherein the generated semantic content information includes identification of one or more types of objects represented by the multiple objects, one or more types of activities that are occurring in the image associated with the multiple objects, or both, and
processing the image data with a second neural network to identify a background and the multiple objects in an image of the plurality of images, to determine a temporal direction for each of the multiple objects based at least in part on motion of the multiple objects identified in the image, the temporal direction describing a 3D motion of each the multiple objects in the image at a point in time, and to generate a separate model for the background and for each of the multiple objects, the model of the background being static and the model of each of the multiple objects including the respective temporal direction for the object; and
generating the 3D object models for the multiple objects based at least in part on the image data and the generated information regarding the identified one or more types of objects, one or more types of activities, or both and on the determined temporal direction of each of the one or more objects.
4. The system of claim 3 , wherein the system is to refine a 3D image utilizing the separate models for the background and each of the multiple objects.
5. A method comprising:
generating a three-dimensional (3D) object model of each of multiple objects in a scene for production of 3D imaging, including:
obtaining image data from a plurality of cameras, the data comprising a plurality of images of the multiple objects captured by the plurality of cameras;
performing processing of the image data to identify and separate each of the multiple objects, including at least the following:
processing the image data with a first neural network to generate semantic content information for one or more of the plurality of images, wherein the generated semantic content information includes identification of one or more types of objects represented by the multiple objects, one or more types of activities that are occurring in the image associated with the multiple objects, or both, and
processing the image data with a second neural network to identify a background and the multiple objects in an image of plurality of images, to determine a temporal direction for each of the multiple objects based at least in part on motion of the multiple objects identified in the image, the temporal direction describing a 3D motion of each of the multiple objects in the image at a point in time, and to generate a separate model for the background and for each of the multiple objects, the model of the background being static and the model of each of the multiple objects including the respective temporal direction for the object; and
generating the 3D object models for the multiple objects based at least in part on the image data and the generated information regarding the identified one or more types of objects, one or more types of activities, or both and on the determined temporal direction of each of the one or more objects.
6. The method of claim 5 , further comprising:
refining a 3D image utilizing the separate models for the background and each of the multiple objects.