碩士 / 國立臺灣海洋大學 / 資訊工程學系 / 106 / Ecological conservation is an important topic that have been learning and discussing for a long time. Many experts and scholars have studied and analyzed that if a certain species disappears, it will affect the extinction of other species, and it may also lead to changes in the ecological environment. Therefore, everyone should make an effort in the conservation of the earth's ecology.
Nowadays, the technology is developed, and everyone can capture the best image quality through the mobile phone. If the relevant technology is combined, the established conservation library data will be put into the relevant software, and the image recognition technology will be used to identify the relevant conservation species. In addition to reminding people not to indiscriminately kill this species, it also allows humans to know the beauty of the earth's ecology and related knowledge, and everyone can do their part for the conservation of the earth's environment.
This study will train and learn the images to be discerned through the powerful recognition capabilities of Deep Learning. We use CAFFE as the learning framework for this study and perform image recognition using Faster R-CNN and YOLOv2 models. Finally, according to the experimental results, the overall recognition rate of YOLOv2 is as high as 84%, which is far better than 72% of Faster R-CNN.
| Identifer | oai:union.ndltd.org:TW/106NTOU5394037 |
| Date | January 2018 |
| Creators | Cheng, Chiu-Hung, 鄭秋紅 |
| Contributors | Hsieh, Jun-Wei, 謝君偉 |
| Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
| Language | zh-TW |
| Detected Language | English |
| Type | 學位論文 ; thesis |
| Format | 45 |
Page generated in 0.0024 seconds