Designing a geometrical three-dimensional modeled object by offloading specific tasks to the GPU : technical

The decision relates to an invention relating to designing a geometrical three-dimensional modeled object. The Examining Division considered the inputting a stroke on a screen plane was known, and neither the visualisation nor the repetition of the method steps with user-modified support achieved a further technical effect.

The applicant argued that computing resources were preserved for constructing the 3D-modelled object as GPU could be used for constructing the visualisation.  The Board disagreed as the claim did not specify that GPU was used.

However, the Auxiliary Request (amended to indicate that certain steps were performed by the GPU) offloaded specific (simple and repetitive) tasks appropriate for the GPU and thus preserved the computing resources of a CPU or memory. Thus, the Board agreed it was technical.

Here are the practical takeaways from the decision: T 1721/20 (Designing a geometrical three-dimensional modeled object/DASSAULT SYSTÈMES) of March 26, 2024 of the Technical Board of Appeal 3.5.07

Key takeaways

The invention

Typically, a process such as designing a car includes a quick, rough sketch to capture the basic idea and then a more detailed, final drawing based on the sketch. The invention lets you draw and design 3D models directly on your computer without needing paper sketches first, making the process faster and more efficient. This is done by discretising and projecting the stroke into a set of points directly on a support in the three-dimensional scene, as well as constructing and visualising the 3D modelled object from these points and support.

 

Is it patentable?

The Examining Division considered The Examining Division considered that inputting a stroke on a screen plane was known, and neither the visualisation nor the repetition of the method steps with user-modified support achieved a further technical effect.

The Board noted that the distinguishing features of claim 1 of the main request over document D1 correspond to features C1, D, D1, E1 and E2, reproduced below:

C1 – projecting the set of points (S130) onto atleast one support in the three-dimensional scene;

D – constructing the three-dimensional modeled object (S140) from the projected set of points and the said at least one support

D1 – by storing together in a file the positions of the projected set of points and data concerning the support;

E1 – computing, for each point of the projected set of points, at least twooffset points (S153), an offset point being a point positioned on said at least one support and spaced from said each point; and

E2 – computing a surface (S155) delineated by the computed at least two offset points, the surface comprising the projected set of points;

Firstly, the Applicant argued that it was a new type of human-machine design:

11.1 The appellant argued that the distinguishing features interacted together and with the other method steps such that a new type of human-machine design interaction process was provided, having the ergonomic advantages of free-hand design, and the system’s responsiveness to the interactions was improved to provide fast visual feedback to the user during the design process of a 3D modelled object.

However, the Board considered the problem :

11.3 However, the board is of the opinion that the claimed invention does not have the aforementioned advantages over the method of document D1, which already provides the same type of ergonomic advantages of free-hand design of three-dimensional models with fast visual feedback during the design process.

Next, the applicant argued that it provided a reduced amount of computing resources for the construction of the 3D-modelled object:

11.5 The appellant argued that distinguishing feature D1 required storing data relative to pixels activated by the stroke. Feature D1 resulted in a reduced amount of computing resources needed for the construction of the 3D modelled object. Moreover, since computing resources did not limit the fluidity and flexibility of the user-machine design interaction, the ergonomics of the new user-machine interaction provided by the claimed invention was improved.

Again, the Board disagreed:

11.6 The board is not convinced that the amount of computing resources needed for the construction of the 3D modelled object is reduced compared with the system of document D1, for example. It is not apparent to the board in which sense the “pixel-based” construction of the 3D modelled object would be more “economical”, or how this would contribute to reducing the amount of computing resources needed for computing the visualisation when compared with document D1.

Finally, the Applicant argued that it provided visual feedback to the user during the design process while preserving computing resources:

1.7 The appellant argued that features E1 and E2 achieved the effect of providing visual feedback to the user during the design process while preserving computing resources. The construction of the visualisation used the stored data relative to the activated pixels to compute the offset points (reference was made to step S153 of Figure 1) and then a tessellation of the surface delineated by these offset points (reference was made to step S155 of Figure 1). This pixel-based construction further reduced the amount of computing resources needed for computing the visualisation in comparison to (the manipulation of) geometric modelled objects and a GPU could be used for constructing the visualisation (step S150), which again was enabled by the storing in step S140. As a result, the speed and fluidity, and thus the ergonomics, of the user-machine design interaction was improved, as lag or delay was reduced.

The Board did not find this argument convincing either:

11.8 The board does not find the appellant’s arguments convincing. Claim 1 does not specify that a GPU is used in the construction of the visualisation. Document D1 also provides visual feedback to the user during the design process (see, for example, Figure 3). The board further notes that claim 1 does not define a tessellation. The subject-matter of claim 1 merely comprises a step of “computing a surface delineated by the computed at least two offset points, the surface comprising the projected set of points” (see feature E2). This computation of the surface is not defined further in claim 1. The board is not convinced that the computation of the “surface delineated by the computed at least two offset points, the surface comprising the projected set of points” requires less computing resources than the system of document D1.

Therefore, it is not apparent from the subject-matter of claim 1 that lag or delay is reduced as compared to the method of document D1.

Therefore, the Board agreed with the Examining Division:

11.10 As stated by the examining division, the effect achieved by distinguishing features C1, D, E1 and E2 is to construct a different visualisation or presentation, for example with construction’s steps towards a “ribbon-like” presentation (instead of a wireframe, for example; see Figure 9 of the application), which is not a technical effect.

Is Auxiliary Request patentable?

The Applicant filed an Auxiliary Request with the following amendments:

Claim 1 of the first auxiliary request differs from claim 1 of the main request in that “coordinates of the support in the 3D scene” are stored (instead of “data concerning the support”) and in that it specifies in feature D1 that the position of the projected set of points and coordinates of the support in the 3D scene are stored in a CAD file, and in that steps E1 and E2 of computing offset points and a surface are performed by a GPU.

The Board acknowledged that the cited document did not disclose GPU in any phase of the generation and visualisation of the object model, but considered the feature of storing in a CAD file obvious.

The applicant argued that it :

14.6 With respect to the first auxiliary request, the appellant maintained essentially the same arguments it had made with respect to the main request (see points 11.1 and 11.2 above). The distinguishing features contributed to preserving computing resources of the CPU and memory resources. The enhanced responsiveness of the system improved the ergonomics in the context of the sketch-based design of claim 1, by providing immediate (real-time) visual feedback to the user.

..

14.8 As the appellant argued, document D1 states that while its mesh construction process takes place in real time, it causes a short pause (or “lag” in the wording of the appellant) of a few seconds when the model becomes complicated (see D1, section 6). The appellant noted that nothing in the section of document D1 entitled “Future Work” (see section 8 of D1) hinted at using the GPU to perform some parts of the construction or to change its mathematics. The construction of D1 used Delaunay triangulation, which was different from the construction of claim 1. In document D1, the steps of finding the chordal axis could not be performed by a GPU since that involved finding centroid points by solving a system of equations using inversion of matrices. It was not obvious to use a GPU to perform this type of computation.

14.9 The appellant argued that the CPU was the default computational unit of the computer. It was designed to handle any computation that the computer might have to perform. The GPU was designed to offload the computational burden on the CPU. It could not perform any computation, as the CPU could, but it could be used for some types of computation (typically “simple and repetitive tasks”), which offloaded the burden on the CPU (the appellant referred here to page 14, lines 21 to 28, of the application as filed). The GPU was faster than the CPU at these types of computation that the GPU could handle because it comprised more cores and parallelised the computations efficiently. Moreover, computing resources of the CPU were saved since the GPU could perform some of the computations in the claimed method, which was what the appellant meant by “less computing resources of the CPU being required”.

14.10 The board considers the problem to be solved by the distinguishing features as being “how to achieve an efficient construction of a visualisation of a modelled object according to a desired geometric form?”.

The Applicant then argued it was not obvious to provide a GPU for certain task:

14.12 The appellant argued that starting from document D1 the skilled person could, or even would, consider using a GPU since the use of a GPU was already known in the field of computer graphics . However, the generic mesh refinement of document D1 did not correspond to the specific computations of using offset points of features E1 and E2. Even if the skilled person had, in search for a solution to the objective technical problem, thought of consulting document D10 and of adding the solution disclosed in document D10 to the method of document D1, they would still not have arrived at the subject-matter of claim 1 of the first auxiliary request. Starting from document D1, there was no reason to modify the computations. Document D1 used rounded strokes. None of the improvements presented in section 8 of document D1 related to solving the “lag” problem such that the visual feedback would be faster and improved. Furthermore, the other documents cited during examination proceedings, including D10, did not disclose the specific computations of features E1 and E2 of claim 1.

14.13 The board finds the appellant’s arguments regarding the offloading to the GPU of the simple and repetitive tasks of features E to E2 convincing. The board notes that step E1 consists in computing, for each point of the projected set of points, at least two offset points (S153), an offset point being a point positioned on the at least one support and spaced from said each point. The computation of the offset points for a projected set of points is simple. It is a task repeated for each projected set of points (which points, as argued by the appellant, can be ordered and labelled by indices). It might also be parallelised. Therefore, it is a specific task which is appropriate for the processing by a GPU. Therefore, the board is convinced that computing resources of a CPU or memory are preserved by the method of claim 1. As the appellant argued, features E, E1 and E2 are especially adapted for a GPU as they constitute simple repetitive tasks. None of the other cited documents describes the distinguishing features. Even if the skilled person starting from document D1 were to consider using a GPU, they would not have modified the model of document D1 in the particular manner of the distinguishing features of claim 1 of the first auxiliary request.

Therefore, the Board agreed that the subject-matter of Auxiliary Request was inventive over the cited prior art.

More information

You can read the full decision here: T 1721/20 (Designing a geometrical three-dimensional modeled object/DASSAULT SYSTÈMES) of March 26, 2024 of the Technical Board of Appeal 3.5.07

Please share this article if you enjoyed it!