Towards Large-Scale Video Video Object Mining
This addresses the challenge of scalable object detection in unlabeled video data for automotive applications, but it is incremental as it builds on existing tracking and mining techniques.
The paper tackles the problem of object mining in large-scale unlabeled automotive videos by leveraging a generic object tracker, resulting in a dataset of over 360,000 automatically mined object tracks from 560,000 frames and a method for novel category discovery and detector learning.
We propose to leverage a generic object tracker in order to perform object mining in large-scale unlabeled videos, captured in a realistic automotive setting. We present a dataset of more than 360'000 automatically mined object tracks from 10+ hours of video data (560'000 frames) and propose a method for automated novel category discovery and detector learning. In addition, we show preliminary results on using the mined tracks for object detector adaptation.