Visual Object Tracking Performance Measures Revisited

ABSTRACT:

The issue of visual following assessment is donning an extensive assortment of execution measures, and generally experiences absence of agreement about which measures ought to be utilized in investigations. This makes the cross-paper tracker correlation troublesome. Moreover, as a few measures might be less successful than others, the following outcomes might be skewed or one-sided toward specific following angles. In this paper, we return to the famous execution measures and tracker execution perceptions and examine them hypothetically and tentatively.

We demonstrate that few measures are identical from the purpose of data they accommodate tracker examination and, essentially, that some are more weak than the others. In view of our investigation, we limit the arrangement of potential measures to just two correlative ones, depicting precision and power, in this way pushing toward homogenization of the tracker assessment procedure. These two measures can be naturally deciphered and imagined and have been utilized by the ongoing visual protest following difficulties as the establishment for the assessment system.

LEAVE A REPLY

Please enter your comment!
Please enter your name here