Table 2 Learning parameters and results of detection
From: Automated harvesting by a dual-arm fruit harvesting robot
Target | Pear (Housui) | Apple (Fuji) | ||
|---|---|---|---|---|
Environment | front-lit | back-lit | front-lit | back-lit |
Architecture | TensorFlow | |||
Net | Mobile Net | |||
Fine Tuning | ssd mobilenet v2 coco | |||
Base learning rate | 0.0001 | |||
batch size | 32 | |||
Learning times | 56197 | 60000 | ||
Detected | 383 | 414 | 405 | 431 |
Not Detected | 15 | 15 | 12 | 14 |
False Detected | 17 | 12 | 7 | 9 |
Precision | 95.8% | 95.5% | 97.1% | 96.9% |
Recall | 95.2% | 97.8% | 98.3% | 98.0% |
F-measure | 0.955 | 0.968 | 0.977 | 0.974 |