Yet Another Computer Vision Index To Datasets (YACVID) - Details

Stand: 2017-05-01 000000m 06:13:21 - Overview

ok839
Attribute Current Content New
Name (Institute + Shorttitle)Malaya Abrupt Motion (MAMo) Dataset 
Description (include details on usage, files and paper references)The Malaya Abrupt Motion (MAMo) dataset is targeted for visual tracking, particularly for abrupt motion tracking. It was collected from publicly accessible dataset used for tracking abrupt motion, as well as animated videos exhibiting objects moving with abrupt motion from the Youtube. It has been used in the paper: M.K. Lim, C.S. Chan, D. Monekosso and P. Remagnino (2014) "Refined Particle Swarm Intelligence Method for Abrupt Motion Tracking", Information Sciences, vol. 283, pp. 267 - 287.


This collection of data comprises various challenging scenarios of abrupt motion, including

1. Rapid motion of small object
- 5 video sequences (TableT1 to TableT5) to test the effectiveness of tracking small object (i.e. table tennis ball) that exhibits fast motion.
- Size of the table tennis ball is very small; about 8x8 pixels to 15x15 pixels for an image resolution of 352 x 240.

2. Switching camera
- 3 video sequences (Youngki, Boxing and Malaya1) to evaluate the effectiveness in tracking both the abrupt and smooth motion.
- The Youngki and Boxing are taken from Kwon and Lee - Wang-Landau Monte Carlo-based Tracking methods for Abrupt Motions, TPAMI2013
- The Malaya1 is created by combining the frames in the Boxing and Youngki sequences in an alternative manner.

3. Partially low-frame rate
- 1 video sequence (Tennis) to test the effectiveness of tracking during low-frame rates.
- It is taken from Kwon and Lee - Wang-Landau Monte Carlo-based Tracking methods for Abrupt Motions, TPAMI2013. The frames herein are downsampled from a video with more than 700 original frames, by keeping one frame in every 25 frames.

4. Inconsistent speed
- 2 videos (Malaya2 - Malaya3) to test the effectiveness of tracking with inconsistent speed movement. Both videos are downloaded from the Youtube.
- Malaya2 is a synthetic video sequence where the ball moves randomly across the video with inconsistent speed.
- Malaya3 is a video sequence a subject juggles a soccer ball in a free-style manner. In this video sequence, the camare is not static, and the background is highly textured.

5. Multiple targets
- 1 video sequence (Malaya4) to evaluate the effectiveness of tracking multiple targets.
- Malaya4 is a synthetic video sequences that consists of 2 simulated balls, moving at random speeds.

All image sequences are in JPEG format and the resolution of each sequence varies (e.g. 360x240, 384x288). A folder containing the ground truth of the object position for each sequence is also provided.

6. Dataset Groundtruth
- Folder name = MAMo Dataset Groundtruth
- Total Files (PDF) = 12

If you use this dataset in your work, you should reference:
M.K. Lim, C.S. Chan, D. Monekosso and P. Remagnino (2014) "Refined Particle Swarm Intelligence Method for Abrupt Motion Tracking", Information Sciences, vol. 283, pp. 267 - 287.

@article{Lim2014,
author = "Mei Kuan Lim and Chee Seng Chan and Dorothy Monekosso and Paolo Remagnino",
title = "Refined particle swarm intelligence method for abrupt motion tracking ",
journal = "Information Sciences",
volume = "283",
pages = "267 - 287",
year = "2014",
issn = "0020-0255",
doi = "http://dx.doi.org/10.1016/j.ins.2014.01.003",
url = "http://www.sciencedirect.com/science/article/pii/S0020025514000085",
URL Linkhttp://cs-chan.com/project3.htm 
Files (#)24 
References (SKIPPED)
Category (SKIPPED)Tracking 
Tags (single words, spaced)visual tracking, abrupt motion tracking 
Last Changed2017-05-01 
Turing (2.12+3.25=?) :-)