Improved Hierarchical Convolutional Features for Robust Visual Object Tracking

Figure 9

Qualitative comparison between the proposed tracker and other representative trackers (KCF [11], HDT [25], Struck [7], and SAMF [22]) on some visual object tracking sequences. (a) Matrix video sequence. (b) Dog video sequence. (c) Singer1 video sequence. (d) Girl2 video sequence. The proposed model provides consistent results in challenging scenarios, such as occlusions, illumination variation, fast motion, and background clutter.

(a)

(b)

(c)

(d)