Related Experiment Video
Updated: Dec 4, 2025
13:36
Author Spotlight: Dendritic Cells Maturation Using Sialidases-Based Enzymatic Treatment of the Cell Surface
Published on: October 20, 2023
4.0K
Testing dataset for head segmentation accuracy for the algorithms in the 'BGSLibrary' v3.0.0 developed by Andrews
Seng Cheong Loke1, Bruce A MacDonald2, Matthew Parsons3
1Faculty of Medicine and Health Sciences, University of Auckland, New Zealand.
Data in Brief
|October 22, 2020
Summary
This dataset evaluates background segmentation algorithms using indoor videos. Results assess accuracy for common factors influencing segmentation, aiding algorithm selection for applications like gamecasting.
Area of Science:
- Computer Vision
- Image Processing
- Algorithm Analysis
Background:
- Background segmentation is crucial for video analysis.
- Existing algorithms face challenges with factors like lighting and motion.
- BGSLibrary offers a C++ wrapper for various segmentation algorithms.
Purpose of the Study:
- To evaluate the accuracy of background segmentation algorithms in BGSLibrary.
- To provide a dataset for testing algorithm performance under various conditions.
- To compare algorithm segmentation accuracy based on common influencing factors.
Main Methods:
- Utilized indoor color-depth video clips featuring a single person.
- Tested BGSLibrary v3.0.0 algorithms on color image data.
- Extracted 'ground truth' from depth data for semi-automatic assessment.
Main Results:
- Segmentation accuracy varied across algorithms and tested factors.
- Depth data provided a reliable basis for 'ground truth'.
- Performance was assessed against criteria relevant to gamecasting and mobile communications.
Conclusions:
- The dataset facilitates objective comparison of background segmentation algorithms.
- Findings inform the selection of appropriate algorithms for specific use cases.
- The study highlights the impact of common factors on segmentation efficacy.