Training data generation apparatus
AssigneeNEC Corporation
InventorTetsuo INOSHITA
Life of this patent
Estimating the term…
Abstract
A selecting unit selects first moving image data and second moving image data from a plurality of frame images composing moving image data. A first generating unit generates first training data that is labeled data relating to a specific recognition target from the frame images composing the first moving image data. A learning unit learns a first model recognizing the specific recognition target by using the first training data. A second generating unit generates second training data that is labeled data relating to the specific recognition target from the frame images composing the second moving image data by using the first model.
Claims17
A training data generation apparatus comprising:
- a memory configured to store instructions;
- and
- a processor configured to execute the instructions to:
- select some frame images from among a plurality of frame images composing first moving image data as second moving image data;
- select, as third moving image data, at least some frame images from among the plurality of frame images composing the first moving image data having not been selected as the second moving image data;
- generate first training data that is labeled data relating to a specific recognition target from the frame images composing the second moving image data;
- learn a first model recognizing the specific recognition target by using the first training data;
- and
- generate, by using the first model, second training data that is labeled data relating to the specific recognition target from the frame images composing the third moving image data.
(canceled)
(canceled)
The training data generation apparatus according to claim wherein the processor is further configured to execute the instructions to determine whether to select all frame images as the third moving image data or some frame images as the third moving image data from among the plurality of frame images composing the first moving image data having not been selected as the second moving image data, based on a total number of the frame images composing the first moving image data.
A training data generation method comprising:
- selecting some frame images from among a plurality of frame images composing first moving image data as second moving image data;
- selecting, as third moving image data, at least some frame images from among the plurality of frame images composing the first moving image data having not been selected as the second moving image data;
- generating first training data that is labeled data relating to a specific recognition target from the frame images composing the second moving image data;
- learning a first model recognizing the specific recognition target by using the first training data;
- and
- generating, by using the first model, second training data that is labeled data relating to the specific recognition target from the frame images composing the third moving image data.
(canceled)
(canceled)
A non-transitory computer-readable recording medium on which a program is recorded, the program comprising instructions for causing a computer to execute:
- a process of selecting some frame images from among a plurality of frame images composing first moving image data as second moving image data;
- a process of selecting, as third moving image data, at least some frame images from among the plurality of frame images composing the first moving image data having not been selected as the second moving image data;
- a process of generating first training data that is labeled data relating to a specific recognition target from the frame images composing the second moving image data;
- a process of learning a first model recognizing the specific recognition target by using the first training data;
- and
- a process of generating, by using the first model, second training data that is labeled data relating to the specific recognition target from the frame images composing the third moving image data.
Dependencies are inferred from the claim text.