# US 11,908,175 B2 · Electronic device training image recognition model and operation method for same > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 11,908,175 B2 - Document ID: US11908175B2 - Status: Granted - Kind: B2 - Application number: 17/693,095 - Filed: 2022-03-11 - Published: 2022-06-23 - Granted: 2024-02-20 - Earliest claimed: 2019-09-11 - Assignees: Samsung Electronics Co., Ltd. - Applicants: Samsung Electronics Co., Ltd. - Inventors: Seowoo Jang; Sangung Yi - Web page: https://patentgrep.com/patent/US11908175B2 ## CPC - G06V 10/10: Image acquisition (document image scanning and transmission H04N1/00; control of digital cameras H04N23/60) - G06V 10/7747: {Organisation of the process, e.g. bagging or boosting} - G06N 3/0464: Convolutional networks [CNN, ConvNet] - G06N 3/08: Learning methods - G06N 3/0895: Weakly supervised learning, e.g. semi-supervised or self-supervised learning - G06N 3/09: Supervised learning - G06T 7/246: using feature-based methods, e.g. the tracking of corners or segments - G06V 10/774: Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting - G06V 10/945: {User interactive design; Environments; Toolboxes} - G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL ## Abstract A method of training an image recognition model includes: generating a virtual dynamic vision sensor (DVS) image using a virtual simulator; generating label information including information about a correct answer to a result of recognition of the DVS image by the image recognition model, with respect to the DVS image; and training the image recognition model by modifying the image recognition model so that a difference between the result of recognition of the DVS image by the image recognition model and the label information is reduced. ## Claims ### Claim 1 A method of training an image recognition model, the method comprising: generating a virtual dynamic vision sensor (DVS) image using a virtual simulator; generating label information comprising information about a correct answer to a result of recognition of the DVS image by the image recognition model with respect to the DVS image; and training the image recognition model by modifying the image recognition model so that a difference between the result of recognition of the DVS image by the image recognition model and the label information is reduced. ### Claim 2 (depends on 1) The method of claim 1, wherein the virtual DVS image is generated based on a virtual environment generated by the virtual simulator, and a virtual object placed in the virtual environment. ### Claim 3 (depends on 2) The method of claim 2, wherein the label information is obtained based on predefined characteristic information with respect to at least one of the virtual environment or the virtual object. ### Claim 4 (depends on 2) The method of claim 2, wherein, in a real environment corresponding to the virtual environment, an object capable of being captured as a real DVS image, which is capable of being recognized by the image recognition model, is placed in the virtual environment as the virtual object. ### Claim 5 (depends on 1) The method of claim 1, wherein the generating of the virtual DVS image comprises: determining at least one camera view point in the virtual environment generated by the virtual simulator; and generating at least one virtual DVS image simultaneously captured from the at least one camera view point, wherein the image recognition model is trained based on the at least one virtual DVS image. ### Claim 6 (depends on 1) The method of claim 1, wherein, a virtual environment is generated by the virtual simulator based on information about a surrounding environment in which a real DVS image capable of being recognized by the image recognition model is capable of being captured, and the virtual DVS image is generated based on the virtual environment. ### Claim 7 (depends on 6) The method of claim 6, wherein, based on the information about the surrounding environment changing by more than a reference value, the virtual DVS image is generated based on the changed information about the surrounding environment. ### Claim 8 (depends on 1) A non-transitory computer-readable recording medium having recorded thereon a program which, when executed by a processor of an electronic device, cause the electronic device to perform operations including the method of claim 1. ### Claim 9 An electronic device configured to train an image recognition model, the electronic device comprising: a memory storing the image recognition model; and at least one processor configured to: generate a virtual dynamic vision sensor (DVS) image using a virtual simulator, generate label information comprising information about a correct answer to a result of recognition of the DVS image by the image recognition model with respect to the DVS image, and train the image recognition model by modifying the image recognition model so that a difference between the result of recognizing the DVS image by the image recognition model and the label information is reduced. ### Claim 10 (depends on 9) The electronic device of claim 9, wherein the virtual DVS image is generated based on a virtual environment generated by the virtual simulator and a virtual object placed in the virtual environment. ### Claim 11 (depends on 10) The method of claim 10, wherein the label information is obtained based on predefined characteristic information with respect to at least one of the virtual environment or the virtual object. ### Claim 12 (depends on 10) The electronic device of claim 10, wherein, in a real environment corresponding to the virtual environment, an object capable of being captured as a real DVS image, which is capable of being recognized by the image recognition model, is placed in the virtual environment as the virtual object. ### Claim 13 (depends on 9) The electronic device of claim 9, wherein the at least one processor is further configured to: determine at least one camera view point in the virtual environment generated by the virtual simulator and generate at least one virtual DVS image simultaneously captured from the at least one camera view point, wherein the image recognition model is trained based on the at least one virtual DVS image. ### Claim 14 (depends on 9) The electronic device of claim 9, wherein, a virtual environment is generated by the virtual simulator based on information about a surrounding environment in which a real DVS image capable of being recognized by the image recognition mode is capable of being captured, and the virtual DVS image is generated based on the virtual environment. ### Claim 15 (depends on 14) The electronic device of claim 14, wherein, based on the information about the surrounding environment changing by more than a reference value, the virtual DVS image is generated based on the changed information about the surrounding environment.