A method for batch generation of three-dimensional models from photographs for virtual production

Abstract

The application discloses a kind of for virtual production photo batch generation three-dimensional model method, comprising the following steps: user submits image alignment task to solving server in local client;Setting load image alignment generation parameter;Setting running instance number, allocate the CPU and GPU resources of each instance, determine task priority;Submit task, and solving server starts image alignment generation work in turn according to task priority;Save image alignment engineering file;Manual inspection alignment data, operator checks the image alignment file that has been completed, confirms whether image alignment file satisfies generation three-dimensional model demand;Client submits the task that needs three-dimensional model and texture generation;Load setting three-dimensional model, texture, export generation attribute;Setting running instance number, allocate the CPU and GPU resources of each instance, for model texture export setting lower running instance number and corresponding CPU and GPU resources, determine task priority.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of virtual production, and particularly relates to a method for generating three-dimensional models in batches from photos for virtual production. BACKGROUND

[0002] Three-dimensional models are required in large quantities in film and television and virtual shooting, but the existing software for generating three-dimensional models based on photos does not have multi-task management functions and multi-instance running functions, and the operator needs to frequently pay attention to the generation progress of the software for generating models from photos. When a task is completed, the operator needs to manually perform image alignment, three-dimensional model, texture, export and other generation setting operations for the next task. However, since the generation time is not a fixed value, it will have different generation times according to the number of imported photos, often causing the operator to be unable to perform generation operations for the next task in time. In order not to affect the overall production progress, the operator needs to frequently pay attention to the generation progress of the software for generating models from photos, and needs to estimate the task completion time, pay attention to and wait for the completion of the task in advance, in order to shorten the idle time between two tasks as much as possible. In addition, the authorization license of each software for generating models from photos can only be used on one computer. The existing solution is basically to estimate the task completion time in advance, pay attention to the task completion progress in advance, wait for the completion of the task, and then perform generation operations for the next task, in order to ensure the overall generation progress. The license can only be purchased at a high price to purchase more authorization licenses. SUMMARY

[0003] In view of the above problems, the present application provides a method for generating three-dimensional models in batches from photos for virtual production, which is used to automatically perform three-dimensional model generation work through remote task submission and multi-instance, solves the discontinuity between tasks, liberates the operator from estimating and paying attention to the completion time in advance, reasonably utilizes the computing server hardware and time resources, and overall improves the generation efficiency.

[0004] To solve the above technical problems, the present application adopts the following technical solution:

[0005] A method for generating three-dimensional models in batches from photos for virtual production, comprising the following steps:

[0006] S101, the user submits an image alignment task to a computing server on a local client; the image alignment task calculates the camera and three-dimensional space management and calculates the basic three-dimensional point cloud in the graph through algorithm calculation, and uses GPU acceleration calculation;

[0007] S102, load image alignment generation parameters, determine the image path, content, and select generation parameters;

[0008] S103, setting the number of running instances, allocating CPU and GPU resources for each instance, and determining the priority of the task;

[0009] S104, submitting the task, and the calculation server starting the image alignment generation work in sequence according to the task priority;

[0010] S105, saving the image alignment engineering file;

[0011] S106, manually checking the alignment data, the operator checks the completed image alignment file, confirms whether the image alignment file meets the generation of three-dimensional model requirements, if there is an error, manually repeating S104, adding image alignment tasks, adjusting the generation attribute, and then performing S105-S106 until the check is passed;

[0012] S107, the client submits a task of generating a three-dimensional model and texture;

[0013] S108, loading and setting the three-dimensional model, texture, and export generation attribute;

[0014] S109, setting the number of running instances, allocating CPU and GPU resources for each instance, setting the number of running instances and corresponding CPU and GPU resources for model texture export, and determining the priority of the task;

[0015] S110, the calculation server starting the automatic running three-dimensional model generation work in sequence according to the task priority;

[0016] S111, saving the three-dimensional model engineering file;

[0017] S112, starting the texture generation content;

[0018] S113, saving the texture file, and the saved texture file covers the three-dimensional model engineering file;

[0019] S114, exporting the three-dimensional model and texture.

[0020] Preferably, further comprising:

[0021] S115, checking the exported three-dimensional model and texture file, confirming whether the three-dimensional model and texture file meets the three-dimensional production requirements; if there is a problem in the check, adjusting the setting generation attribute, then continuing steps S112-S114 until the check is passed and the task is completed.

[0022] Preferably, the photo scanning software is installed in the calculation server in the machine room, a cluster management server is established, and the client submits the model map generation and export tasks in different stages to the cluster management server, the cluster management server allocates tasks and controls priorities, allocates CPU and GPU hardware resources of the calculation server, and controls the number of virtualization software instances at the same time.

[0023] Preferably, the calculation server is provided with multiple high-performance GPUs, and the calculation server and the cluster management server and the client computer are in the same internal network.

[0024] Preferably, the data is stored in a network storage with access interfaces set on the client and the calculation server.

[0025] The application has the beneficial effects of improving parallel computing capability, automatically allocating and calculating tasks in the calculation server according to priorities, reasonably utilizing hardware performance, and improving overall efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 The figure is a step flow chart of the method for generating three-dimensional models in batches from photos for virtual production according to the embodiment of the application. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the application will be clearly and completely described with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are some but not all of the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the application.

[0028] Referring to Figure 1 The figure is a step flow chart of the method for generating three-dimensional models in batches from photos for virtual production according to the embodiment of the application, including the following steps:

[0029] S101, a user submits an image alignment task to a calculation server in a local client; the image alignment task calculates and inversely calculates basic three-dimensional point clouds in a camera and a three-dimensional space management and calculation graph from photos by using an algorithm, and uses GPU acceleration calculation;

[0030] S102, loading image alignment generation parameters are set, an image path is determined, content is selected, and generation parameters are selected;

[0031] S103, the number of running instances is set, CPU and GPU resources of each instance are allocated, and task priorities are determined;

[0032] S104, a task is submitted, and the calculation server starts image alignment generation work in sequence according to the task priorities.

[0033] S105, save the image alignment engineering file;

[0034] S106, manually check the alignment data, the operator checks the completed image alignment file, confirms whether the image alignment file meets the generation of three-dimensional model requirements, if there is an error, manually repeats S104, adds an image alignment task, adjusts the generation attribute, and then performs in the manner of S105-S106 until the check is passed;

[0035] S107, the client submits a task requiring three-dimensional model and texture generation;

[0036] S108, load the setting three-dimensional model, texture, export generation attribute;

[0037] S109, set the number of running instances, allocate CPU and GPU resources for each instance, and submit the running instance number and corresponding CPU and GPU resources for the model texture export setting, determine the task priority;

[0038] S110, the solving server will start the automatic running three-dimensional model generation work in turn according to the task priority;

[0039] S111, save the three-dimensional model engineering file;

[0040] S112, start the texture generation content;

[0041] S113, save the texture file, and the saved texture file covers the three-dimensional model engineering file;

[0042] S114, export the three-dimensional model and the texture.

[0043] The above setting of the method for batch generating three-dimensional models from photos for virtual production solves the task and task interruption, liberates the operator from pre-estimating attention and waiting for completion time, reasonably utilizes the solving server hardware and time resources, and overall improves the generation efficiency.

[0044] Further, the method for batch generating three-dimensional models from photos for virtual production according to another embodiment of the application, based on S101 to S114, further comprises:

[0045] S115, check the exported three-dimensional model and texture file to confirm whether the three-dimensional model and texture file meets the three-dimensional production requirements; if there is a problem in the check, adjust the generation attribute setting, then continue with steps S112-S114 until the check is passed and the task is completed.

[0046] In an embodiment of the present application, the photo scanning software is installed in the calculation server in the machine room, a cluster management server is established, and the model map generation and export tasks in different stages are submitted to the cluster management server by the client, the cluster management server allocates tasks and controls priority, allocates CPU and GPU hardware resources of the calculation server, and controls the number of virtualization software instances at the same time. The photo scanning software can be, for example, Capture Reality.

[0047] In an embodiment of the present application, the calculation server is provided with multiple high-performance GPUs to ensure that the calculation server and the cluster management server and the client computer are in the same internal network.

[0048] In an embodiment of the present application, the data is stored in a NAS network storage with access interfaces set on the client and the calculation server.

[0049] It should be understood that the example embodiments described herein are illustrative and not restrictive. Although one or more embodiments of the present application are described in conjunction with the attached figures, it should be understood that various changes in form and detail can be made without departing from the spirit and scope of the application as defined by the appended claims.

Claims

1. A method for batch production of three-dimensional models from photographs for virtual production, characterized in that, Comprising the following steps: S101, the user submits an image alignment task to the calculation server at the local client; The image alignment task is to calculate the camera and three-dimensional space management and calculation graph of the basic three-dimensional point cloud by algorithm, using GPU accelerated calculation; S102, set up the image alignment generation parameters, determine the image path, content, select the generation parameters; S103, set the number of running instances, allocate CPU and GPU resources for each instance, and determine the task priority; S104, submit the task, and the calculation server starts the image alignment generation work in turn according to the task priority; S105, save the image alignment project file; S106, manually check the alignment data, the operator checks the completed image alignment file, confirms whether the image alignment file meets the generation of three-dimensional model requirements, if there is an error, manually repeat S104, add image alignment task, adjust the generation attribute, and then perform S105-S106 until the check is passed; S107, the client submits a task of generating three-dimensional model and texture; S108, load the three-dimensional model, texture, and export generation attribute settings; S109, set the number of running instances, allocate CPU and GPU resources for each instance, set the number of running instances and corresponding CPU and GPU resources for model texture export, and determine the task priority; S110, the calculation server starts the three-dimensional model generation work in turn according to the task priority; S111, save the three-dimensional model project file; S112, start the texture generation content; S113, save the texture file, which covers the three-dimensional model project file; S114, export the three-dimensional model and texture.

2. The method for batch generating three-dimensional models from photographs for virtual prototyping of claim 1, wherein, Further comprising: S115, check the exported three-dimensional model and texture file to confirm whether the three-dimensional model and texture file meets the three-dimensional production requirements; if there is a problem, adjust the generation attribute settings, then continue steps S112-S114 until the check is passed and the task is completed.

3. The method for batch production of three-dimensional models from photographs for virtual prototyping according to claim 1 or 2, wherein, Install the photo scanning software in the calculation server in the computer room, establish a cluster management server, submit the model mapping generation and export tasks at different stages to the cluster management server at the client, allocate tasks and control priority by the cluster management server, allocate CPU and GPU hardware resources of the calculation server, and control the number of virtualization software instances at the same time.

4. The method for batch generating three-dimensional models from photographs for virtual prototyping of claim 3, wherein, The calculation server is provided with multiple high-performance GPUs to ensure that the calculation server and the cluster management server, and the client computer are in the same internal network.

5. The method for batch production of three-dimensional models from photographs for virtual prototyping of claim 1 or 2, wherein, The data is stored in the network storage with access interface set in the client and the calculation server.

Images