Using Deep Learning in The Development of A Welding Practice Training Support System
| Autor: | Pham-Tuan-Anh Phung, Vu-Hoang Tran, Ton-Nghia Huynh, Ngoc-Thien Tran |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
Backbone network
Computer science business.industry Deep learning ComputerApplications_COMPUTERSINOTHERSYSTEMS Welding Machine learning computer.software_genre Training (civil) law.invention Development (topology) law Completeness (order theory) Support system Artificial intelligence Architecture business computer |
| Zdroj: | ICSSE |
| Popis: | Along with the strong development of the industrial revolution 4.0 in many fields, the artificial intelligence systems supporting learners have also received more attention. In this paper, we propose a welding practice training support system using artificial intelligence. The system is specially designed to be able to learn from the challenging dataset, that is small but contains many variations of weld, to classify student’s weld based on completeness of weld. To handle the multi-variant of weld defects, we propose cSTN to not only allow the network to adaptively capture the weld defect regions but also make the decision based on both local and global perspectives. Besides, to deal with the insufficient data problem, we propose to slightly modify the DenseNet architecture as our backbone network. Experimental results indicated that the proposed system still achieved good performance despite only training on a small dataset. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|