The experiements of re-detection with confidence score about long term tracking.
The confidence score is the biggest score to select the best location in each frame, which can be the basis of judge whether the target is lost. We can use the single confidence score or the sequenced confidence score to evaluate the tracking process.
Mixformer is a excellent short term trackers and Unicorn is a excellent glocal tracker. Our purpose aims to modify Mixformer to a long object trackers. So we want to add a re-detection mechanism. When the mechanism judge the tracking is failed, the running tracker will exchange to the global tracker to find the lost target object.
In this experiments, the re-detection mechanism exploit the confidence score to judge whether the target is lost.
Lasot is the benchmark of long term object tracking and average confidence score of mixformer in the lasot is 0.58 when IOU is zero.
The experiments are as follows:
lasot | Success | Precision | Norm Precision | Mixformer-base | 0.711 | 0.757 |0.740 | M_U_sc_30 | 0.713 | 0.761 |0.743 | M_U_sc_58 | 0.711 | 0.758 |0.742 | M_U_sc_sq_10_30 | 0.718 | 0.764 |0.749 | M_U_sc_sq_20_30 | 0.719 | 0.766 |0.749 | M_U_sc_sq_50_30 | 0.718 | 0.765 |0.747 | M_U_sc_sq_100_30 | 0.715 | 0.761 |0.745 | M_U_sc_sq_200_30 | 0.714 | 0.760 |0.743 | M_U_sc_psq_10_30 | 0.718 | 0.764 |0.748 | M_U_sc_psq_20_30 | 0.718 | 0.765 |0.748 | M_U_sc_sq_10_58 | 0.715 | 0.762 |0.745 | M_U_sc_sq_200_58 | 0.715 | 0.763 |0.745 |
Where M_U_sc_30 is the tracker using score under 0.3 and M_U_sc_58 is the tracker using score under 0.58. M_U_sc_sq_10_30 means that average ten confidence scores under 0.3. And Psq uses penalty.
The 0.3 may be more effective than 0.58. The best length of sequence may be 20. The penaly seems not effective. It’s too cumbersome to conduct more experiments on hyperparameters
In summary, we can see that exploiting confidence score to re-detect works but is not obvious. So in the next step, we can conduct experiemts on the similarity of templates and tracking object.
