Author Affiliations
1School of Electronic Countermeasure, National University of Defense Technology, Hefei 230037, China2State Key Laboratory of Pulsed Power Laser Technology, Hefei 230037, Chinashow less
Fig. 1. Classification response map of SiamRPN
Fig. 2. Schematic diagram of the two-stage tracking method based on Siamese network
Fig. 3. Features extracted by different networks
Fig. 4. Comparative analysis of network outputs after applying correlation filter modulation
Fig. 5. Schematic diagram of the verification network
Fig. 6. Comparison of classification distances of the inter-object and intra-object
Fig. 7. Comparison of classification distance between different sampling groups
Fig. 8. Evaluation on the OTB100 benchmark: Precision plot and success plot
Fig. 9. AR scores in baseline test condition
Fig. 10. Overlap rate curves in unsupervised condition
Fig. 11. Test results on the UAV123 dataset: Precision plot and success plot
Fig. 12. Qualitative evaluation of challenging sequences
Fig. 13. Ability to deal with the similar interference of the two-stage Siamese network
Fig. 14. Comparison of regression precision of the two-stage Siamese network
Fig. 15. Running time test of each module
| Baseline | | Unsupervised | A-R rank | EAO | | Overlap | Speed | Overlap | Failures | EAO | | AUC | Normalized | FPS | Ours | 0.601 1 | 14.515 9 | 0.383 3 | | 0.533 9 | 3.496 1 | 20.245 1 | LADCF | 0.491 1 | 9.925 3 | 0.3811 | 0.418 2 | 0.123 0 | 0.557 3 | MFT | 0.491 9 | 10.766 2 | 0.379 4 | 0.391 7 | 0.194 5 | 0.623 2 | DaSiamRPN | 0.569 1 | 18.441 5 | 0.378 5 | 0.468 4 | 17.818 3 | 64.414 3 | UPDT | 0.515 4 | 11.417 2 | 0.371 9 | 0.444 4 | 0.088 4 | 0.469 7 | RCO | 0.498 9 | 10.700 4 | 0.371 1 | 0.383 0 | 0.204 6 | 0.720 3 | SiamRPN | 0.591 5 | 19.632 5 | 0.369 1 | 0.456 8 | 20.342 6 | 86.784 3 | DRT | 0.495 8 | 13.947 6 | 0.349 0 | 0.419 1 | 0.123 7 | 0.456 8 | DeepSTRCF | 0.506 2 | 14.548 6 | 0.338 3 | 0.433 3 | 0.560 5 | 3.114 4 | CPT | 0.488 8 | 16.620 7 | 0.332 1 | 0.375 7 | 0.877 1 | 5.184 2 | SA_Siam_R | 0.544 4 | 16.403 0 | 0.331 1 | 0.425 0 | 6.776 1 | 32.364 4 | DLSTpp | 0.529 7 | 14.937 4 | 0.321 3 | 0.497 8 | 1.293 0 | 8.175 9 |
|
Table 1. Evaluation results on the VOT benchmark