# US 2021/0241442 A1 · Calculation method, calculation apparatus, and computer-readable recording medium > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 2021/0241442 A1 - Document ID: US20210241442A1 - Status: Published application - Kind: A1 - Application number: 17/129,730 - Filed: 2020-12-21 - Published: 2021-08-05 - Earliest claimed: 2020-01-30 - Assignees: Fujitsu Limited - Applicants: Fujitsu Limited - Inventors: Akira Sakai - Web page: https://patentgrep.com/patent/US20210241442A1 ## CPC - G06T 7/0002: {Inspection of images, e.g. flaw detection} - G06T 2207/10016: Video; Image sequence - G06T 2207/20076: Probabilistic image processing - G06T 2207/20084: Artificial neural networks [ANN] - G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL ## Abstract A non-transitory computer-readable recording medium stores therein a calculation program that causes a computer to execute a process including setting, with regard to each of a plurality of frame images included in moving image data on a captured subject, a time interval region having a predetermined time width in time sequence for detection map information relating a detection result of each of types of feature values to a time sequence of the frame images, acquiring, with regard to each of a plurality of positions in the time interval region, a distribution state of the detection result in the time interval region in a case where a position of the set time interval region is changed in the time sequence in the detection map information, and calculating a distributed representation indicating a set of points obtained by projecting the distribution state in each of the positions onto a low-dimensional space. ## Claims ### Claim 1 A non-transitory computer-readable recording medium storing therein a calculation program that causes a computer to execute a process, the process comprising: setting, with regard to each of a plurality of frame images included in moving image data on a captured subject, a time interval region including a predetermined time width in time sequence for detection map information relating a detection result of each of types of feature values to a time sequence of the frame images; acquiring, with regard to each of a plurality of positions in the time interval region, a distribution state of the detection result in the time interval region in a case where a position of the set time interval region is changed in the time sequence in the detection map information; calculating a distributed representation indicating a set of points obtained by projecting the distribution state in each of the positions onto a low-dimensional space; and determining, based on a comparison between the acquired distributed representation and the distributed representation obtained from determination target moving image data on the captured subject, whether the determination target moving image data is data on a captured subject in a faulty state. ### Claim 2 (depends on 1) The non-transitory computer-readable recording medium according to claim 1, the process further including detecting, with regard to each of the frame images, a degree of certainty that is an index corresponding to a level of probability that each of predetermined objects is present, wherein the acquiring includes acquiring the degree of certainty as the distribution state. ### Claim 3 (depends on 2) The non-transitory computer-readable recording medium according to claim 2, wherein the detecting includes detecting a size and a position of a rectangular region which is present in the frame image and in which presence probability of the object is equal to or more than a predetermined value, and the acquiring includes acquiring, as the distribution state, the degree of certainty and the size and the position of the rectangular region. ### Claim 4 (depends on 1) The non-transitory computer-readable recording medium according to claim 1, wherein the calculating includes inputting the distribution state to an autoencoder including an intermediate layer made up of two units to calculate a distributed representation indicating a set of points projected onto a two-dimensional space. ### Claim 5 (depends on 4) The non-transitory computer-readable recording medium according to claim 4, the process further including displaying, on a screen, a trajectory of a point projected onto the two-dimensional space in the time sequence. ### Claim 6 (depends on 1) The non-transitory computer-readable recording medium according to claim 1, wherein the detection map information is a two-dimensional map in which the detection result is mapped on a two-dimensional plane whose first axis is one of the types of feature values and whose second axis is the time sequence. ### Claim 7 A calculation method executed by a computer, the calculation method comprising: setting, with regard to each of a plurality of frame images included in moving image data on a captured subject, a time interval region including a predetermined time width in time sequence for detection map information relating a detection result of each of types of feature values to a time sequence of the frame image; acquiring, with regard to each of a plurality of positions in the time interval region, a distribution state of the detection result in the time interval region in a case where a position of the set time interval region is changed in the time sequence in the detection map information; calculating a distributed representation indicating a set of points obtained by projecting the distribution state in each of the positions onto a low-dimensional space; and determining, based on a comparison between the acquired distributed representation and the distributed representation obtained from determination target moving image data on the captured subject, whether the determination target moving image data is data on a captured subject in a faulty state. ### Claim 8 A calculation apparatus comprising: a memory; and a processor coupled to the memory and configured to: set, with regard to each of a plurality of frame images included in moving image data on a captured subject, a time interval region including a predetermined time width in time sequence for detection map information relating a detection result of each of types of feature values to a time sequence of the frame image, acquire, with regard to each of a plurality of positions in the time interval region, a distribution state of the detection result in the time interval region in a case where a position of the time interval region set at the setting is changed in the time sequence in the detection map information, calculate a distributed representation indicating a set of points obtained by projecting the distribution state in each of the positions onto a low-dimensional space, and determine, based on a comparison between the acquired distributed representation and the distributed representation obtained from determination target moving image data on the captured subject, whether the determination target moving image data is data on a captured subject in a faulty state.