Table 1 Performance of different YOLO algorithms for processing the fish training dataset on the industrial computer platform in this study
From: Automated under-ice monitoring system for fish detection in Central Arctic Ocean
Model | Mean average precision (mAP@50) | Frames per second (FPS) | Total processing time (s) | GPU peak power consumption (W) | Total power consumption per task (J) |
|---|---|---|---|---|---|
YOLOv5 | 97.0% | 28 | 41.2 | 48–52 | 512 |
YOLOv7 | 97.1% | 25 | 47.8 | 55–60 | 592 |
YOLOv8 | 98.4% | 21 | 52.4 | 62–66 | 685 |