Исследование сверточной нейронной сети небольшой архитектуры для распознавания жестов
| Parent link: | Современные технологии, экономика и образование: сборник трудов Всероссийской научно-методической конференции, г. Томск, 27-29 декабря 2019 г./ Национальный исследовательский Томский политехнический университет ; под ред. А. Г. Фефеловой, Е. А. Покровской, И. О. Болотиной [и др.]. [С. 157-159].— , 2019 |
|---|---|
| Glavni avtor: | |
| Korporativna značnica: | |
| Drugi avtorji: | |
| Izvleček: | Заглавие с титульного экрана Currently, more research is aimed at solving problems using computer vision and artificial intelligence. Most frequent are solutions and approaches using gesture recognition based on infrared sensors or neural networks. The relevance of the subject matter is due to the possibility of applying the proposed approach for managing the operation of objects without tactile contact and voice identification of commands, as well as its simplicity from the point of view of the end-user. This paper proposes a proprietary convolutional neural network architecture to solve gesture classification. The accuracy of the network operation was evaluated depending on the distance between the camera and the hand, as well as depending on the complexity of the gesture. |
| Izdano: |
2019
|
| Serija: | Современные образовательные технологии и тенденции развития инженерного образования в России |
| Teme: | |
| Online dostop: | http://earchive.tpu.ru/handle/11683/58182 |
| Format: | Elektronski Book Chapter |
| KOHA link: | https://koha.lib.tpu.ru/cgi-bin/koha/opac-detail.pl?biblionumber=631023 |
| Izvleček: | Заглавие с титульного экрана Currently, more research is aimed at solving problems using computer vision and artificial intelligence. Most frequent are solutions and approaches using gesture recognition based on infrared sensors or neural networks. The relevance of the subject matter is due to the possibility of applying the proposed approach for managing the operation of objects without tactile contact and voice identification of commands, as well as its simplicity from the point of view of the end-user. This paper proposes a proprietary convolutional neural network architecture to solve gesture classification. The accuracy of the network operation was evaluated depending on the distance between the camera and the hand, as well as depending on the complexity of the gesture. |
|---|