講演抄録/キーワード |
講演名 |
2019-10-10 14:50
Segnet and U-Net Implementations for Water Hyacinth Semantic Segmentation in Thailand ○Supatta Viriyavisuthisakul・Parinya Sanguansat(PIM)・Toshihiko Yamasaki(UTokyo) MVE2019-25 |
抄録 |
(和) |
Water Hyacinth is an aquatic weed that can spread very quickly. Normally, it can be found in a dam or river. Water Hyacinth is one of the main problems in irrigation management, water transportation, and agriculture. That makes it is one of the national problems, especially Thailand. To plan the solution for this problem, quantitative measurement of Water Hyacinth in many areas is required. Semantic segmentation with deep learning is very accurate now, and it can be used in this task. In this paper, the semantic segmentation is applied to segment the Water Hyacinth in the images. Segnet and U-Net were compared. Both of them use convolutional layers but in different architectures. With our limited resources, we found that U-Net is more suitable in this task than Segnet in both computational time and performance. |
(英) |
Water Hyacinth is an aquatic weed that can spread very quickly. Normally, it can be found in a dam or river. Water Hyacinth is one of the main problems in irrigation management, water transportation, and agriculture. That makes it is one of the national problems, especially Thailand. To plan the solution for this problem, quantitative measurement of Water Hyacinth in many areas is required. Semantic segmentation with deep learning is very accurate now, and it can be used in this task. In this paper, the semantic segmentation is applied to segment the Water Hyacinth in the images. Segnet and U-Net were compared. Both of them use convolutional layers but in different architectures. With our limited resources, we found that U-Net is more suitable in this task than Segnet in both computational time and performance. |
キーワード |
(和) |
Water Hyacinth / Semantic Segmentation / Segnet / U-Net / / / / |
(英) |
Water Hyacinth / Semantic Segmentation / Segnet / U-Net / / / / |
文献情報 |
信学技報, vol. 119, no. 222, MVE2019-25, pp. 9-12, 2019年10月. |
資料番号 |
MVE2019-25 |
発行日 |
2019-10-03 (MVE) |
ISSN |
Print edition: ISSN 0913-5685 Online edition: ISSN 2432-6380 |
著作権に ついて |
技術研究報告に掲載された論文の著作権は電子情報通信学会に帰属します.(許諾番号:10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
PDFダウンロード |
MVE2019-25 |