To search, Click below search items. |
| All Published Papers Search Service |
|
|
|
Title |
Fast and Accurate Fish Detection Design with Deep Learning YOLO-v3 Model and Transfer Learning |
|
Author |
Kazim Raza and Zhoushan, Zhejiang, China |
| Citation |
Vol. 26 No. 1 pp. 35-42 |
|
Abstract |
Object Detection is one of the difficult computer vision problems with countless applications. We proposed a real-time object detection algorithms based on YOLOv3 for detecting low accuracy objects and slow speed of detection. The demand for monitoring the marine ecosystem is increasing day by day for a vigorous automated system, which could be beneficial for all of the researchers in order to collect information about marine life. This proposed work mainly approached the computer vision technique to detect and classify marine life. Most such systems are already developed that totally based on CNNs where a large amount of training data required. In this paper, we performed object detection on four fish species custom datasets by applying YOLOv3 architecture, where we got 87.56% mAP (mean average precision). We also worked on improving the YOLOv3 baseline model with the help of a novel transfer learning technique, and improvement in loss function to improve the model performance. Moreover, comparing to the experimental analysis of the original YOLOv3 model with the improved one, we observed the mAP increased from 87.56% to 91.30. It showed that improved version outperforms than the original YOLOv3 model. |
|
Keywords |
Deep learning, computer vision, Transfer Learning, Improved YOLOv3, Anchor Box, Custom dataset |
|
URL |
