STD: A Stereo Tracking Dataset for evaluating binocular tracking algorithms
| Autor: | Zheng Zhu, Wei Zou, Qingbin Wang, Feng Zhang |
|---|---|
| Rok vydání: | 2016 |
| Předmět: |
0209 industrial biotechnology
Ground truth business.industry Computer science ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Stereoscopy Mobile robot 02 engineering and technology Object (computer science) Tracking (particle physics) Stereo tracking law.invention 020901 industrial engineering & automation law Scalability 0202 electrical engineering electronic engineering information engineering 020201 artificial intelligence & image processing Computer vision Artificial intelligence Scale (map) business Algorithm |
| Zdroj: | ROBIO |
| DOI: | 10.1109/robio.2016.7866659 |
| Popis: | In this paper, a Stereo Tracking Dataset is proposed for evaluating binocular tracking algorithms. The dataset contains stereoscopic videos which are collected by our mobile platform in different scenarios and videos that are available publicly. All sequences are carefully synchronized and rectified, and the ground truth of object is annotated by authors. Both raw and processed sequences are provided in the dataset. We also develop a Scalable and Occlusion-aware Multi-cues Correlation Filter Tracker (SOMCFT) and evaluate it on the STD. The SOMCFT framework fuses different clues in confidence map level and uses depth information to handle scale changes and occlusion. Quantitative evaluation on STD demonstrates effectiveness of the proposed dataset. All data, including stereo image pairs, calibrations, annotations and attributes, are available for research purposes and comparative evaluation on https://github.com/zhengzhugithub/StereoTracking. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|