Cannot See the Forest for the Trees: Aggregating Multiple Viewpoints to Better Classify Objects in Videos
This work addresses the challenge of improving classification accuracy in videos for long-tailed object tracking, which is incremental as it builds on existing trackers with a plug-and-play method.
The paper tackles the problem of inconsistent classification in long-tailed object tracking by introducing a set classifier that aggregates multiple viewpoints within tracklets, achieving state-of-the-art results of 19.9% and 15.7% TrackAP_50 on TAO validation and test sets.
Recently, both long-tailed recognition and object tracking have made great advances individually. TAO benchmark presented a mixture of the two, long-tailed object tracking, in order to further reflect the aspect of the real-world. To date, existing solutions have adopted detectors showing robustness in long-tailed distributions, which derive per-frame results. Then, they used tracking algorithms that combine the temporally independent detections to finalize tracklets. However, as the approaches did not take temporal changes in scenes into account, inconsistent classification results in videos led to low overall performance. In this paper, we present a set classifier that improves accuracy of classifying tracklets by aggregating information from multiple viewpoints contained in a tracklet. To cope with sparse annotations in videos, we further propose augmentation of tracklets that can maximize data efficiency. The set classifier is plug-and-playable to existing object trackers, and highly improves the performance of long-tailed object tracking. By simply attaching our method to QDTrack on top of ResNet-101, we achieve the new state-of-the-art, 19.9% and 15.7% TrackAP_50 on TAO validation and test sets, respectively.