Table 8.
Precision (P) and recall (R) at IoU = 0.5 of different CNN marker detectors with Profile Checker algorithms.
| Morning | Afternoon | Evening | Entire Dataset | |||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Far | Close | Far | Close | Far | close | Far | Close | Far +close | ||||||||||
| P | R | P | R | P | R | P | R | P | R | P | R | P | R | P | R | P | R | |
| lightDenseYOLO +Profile Checker v1 | 0.97 | 0.96 | 0.96 | 0.95 | 0.96 | 0.97 | 0.96 | 0.98 | 0.95 | 0.96 | 0.96 | 0.96 | 0.96 | 0.96 | 0.96 | 0.96 | 0.96 | 0.96 |
| lightDenseYOLO +Profile Checker v2 | 0.99 | 0.99 | 0.98 | 0.99 | 0.98 | 0.98 | 0.99 | 0.98 | 0.98 | 0.99 | 0.99 | 0.99 | 0.98 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 |
| YOLO v2 + Profile Checker v2 | 0.98 | 0.97 | 0.99 | 0.99 | 0.98 | 0.99 | 0.99 | 0.98 | 0.98 | 0.98 | 0.99 | 0.99 | 0.98 | 0.98 | 0.99 | 0.99 | 0.99 | 0.98 |
| Faster R-CNN + Profile Checker v2 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 |
| MobileNets-SSD + Profile Checker v2 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 | 0.99 |