A New Trajectory Based Motion Segmentation Benchmark Dataset (UdG-MS15)
Motion segmentation (MS) is an essential step in video analysis. Its quantitative and qualitative evaluation is largely dependent on the dataset used for testing. Although there are publicly available datasets such as Hopkins and FBMS, they have limitations in terms of number of motions, partial/complete occlusion, stopping motion, sequence length, and real life natural sequences. Due to these limitations, many recent proposals have reached nearly zero misclassification, especially for Hopkins, which leaves no room for quantitatively differentiating among proposals. In this paper, we present a new challenging trajectory based MS dataset of 15 sequences, where number of motions and sequence length have been largely increased as compared to the state of the art. An effort has been made to include all forms of distortions that are present in real life scenes. As a starting point, a preliminary benchmark evaluation using a recent and well known state of the art algorithm has been provided for this dataset.
KeywordsMotion segmentation Tracking Trajectory Benchmark Dataset
This work has been supported by the FP7-ICT-2011-7 project PANDORA-Persistent Autonomy through Learning, Adaptation, Observation and Re-planning (Ref 288273) funded by the European Commission and the project RAIMON-Autonomous Underwater Robot for Marine Fish Farms Inspection and Monitoring (Ref CTM2011-29691-C02-02) funded by the Ministry of Economy and Competitiveness of the Spanish Government. Muhammad Habib Mahmood is supported by an FI grant associated with the RAIMON project.
- 1.Liu, G., Yan, S.: Latent low-rank representation for subspace segmentation and feature extraction. In: ICCV, pp. 1615–1622 (2011)Google Scholar
- 5.Zappella, L., Del Bue, A., Lladó, X., Salvi, J.: Joint estimation of segmentation and structure from motion. CVIU 117, 113–129 (2013)Google Scholar
- 11.Ochs, P., Brox, T.: Higher order motion models and spectral clustering. In: CVPR, pp. 614–621 (2012)Google Scholar
- 12.Tron, R., Vidal, R.: A benchmark for the comparison of 3-D motion segmentation algorithms. In: CVPR, pp. 1–8 (2007)Google Scholar
- 17.Badrinarayanan, V., Galasso, F., Cipolla, R.: Label propagation in video sequences. In: CVPR, pp. 3265–3272 (2010)Google Scholar
- 18.Budvytis, I., Badrinarayanan, V., Cipolla, R.: Label propagation in complex video sequences using semi-supervised learning. BMVC 2257, 2258–2259 (2010)Google Scholar