An acquisition unit acquires frames constituting an image. An operation specification unit specifies operation details based on the image of the frames by way of machine learning using a learning model. A period specification unit corrects the operation details associated with the image of the frames based on time series relationship between a first period and a second period of periods included in the image and based on a predetermined operation order of a plurality of operation details and to specify a period of each operation corresponding to the operation order, the first period corresponding to continuous frames in which a first operation detail is specified by the operation specification unit, the second period being different from the first period and corresponding to continuous frames in which a second operation detail is specified.