US 12,249,134 B2
Visualization method, program for the same, visualization device, and discrimination device having the same
Takayuki Ishiguro, Nagoya (JP); and Hitoshi Hoshino, Nagoya (JP)
Assigned to ROXY CORP., Aichi (JP)
Appl. No. 17/907,135
Filed by ROXY CORP., Aichi (JP)
PCT Filed Apr. 1, 2021, PCT No. PCT/JP2021/014131
§ 371(c)(1), (2) Date Sep. 23, 2022,
PCT Pub. No. WO2021/210414, PCT Pub. Date Oct. 21, 2021.
Claims priority of application No. 2020-073810 (JP), filed on Apr. 17, 2020.
Prior Publication US 2023/0103374 A1, Apr. 6, 2023
Int. Cl. G06V 10/00 (2022.01); G06V 10/77 (2022.01); G06V 10/771 (2022.01); G06V 10/82 (2022.01); G06V 10/86 (2022.01)
CPC G06V 10/86 (2022.01) [G06V 10/771 (2022.01); G06V 10/7715 (2022.01); G06V 10/82 (2022.01)] 25 Claims
OG exemplary drawing
 
1. A visualization method for visualizing a trained state of a model that has been trained with a plurality of sample data, the method including:
(a) in order to generate the trained model, extracting first multi-dimensional feature vectors of each of the plurality of sample data by a feature extractor;
(b) converting the extracted first multi-dimensional feature vectors to second multi-dimensional feature vectors of a lower dimension than the first multi-dimensional feature vectors by a feature converter that uses a plurality of fully connected layers;
(c) graphing the converted second multi-dimensional feature vectors of each of the plurality of sample data; and
(d) displaying a plurality of generated graphs in a first section of a display unit.