Week 22 - 28 June
- Review of the provisory dissertation and respective modification
Week 9 - 21 June
- Write dissertation
Week 8 - 14 June
- Write dissertation
Week 1 - 7 June
- Write dissertation
Week 25 - 31 May
- Obtain results from both algorithms under the BoBoT dataset
Week 18 - 24 May
- Continuation with the improvements in the Android application
- Upgrade of the benchmark software to fit the evaluation methods required
Week 11 - 17 May
- Application updated with new algorithm named CMT
- Comparison between CMT and TLD performance using benchmark software created previous week
Week 4 - 10 May
- Finished developing desktop software and continuing with benchmark application (C++ & OpenCV)
3rd Meeting - 7th of May
- Discussion about work already done and overview of key aspects
- Debate what methods should be use to benchmark software
Week 27 April - 3 May
- Improvement and continuing development of application features started the previous week
- Start developing test bench software to benchmark algorithm's correctness (C++ & OpenCV)
Week 20 - 26 April
- Implementation of TLD tracking algorithm in Android environment (Android NDK, OpenCV, C++ & Java):
- Demo application created (Video)
- Creation of new features started:
- Initialize tracker using body detector (HOG and Haars cascade classifiers) instead of choosing object's ROI
Week 13 - 19 April
- Study TLD (Predator) algorithm and OpenTLD library
- Start implementation of TLD tracking algorithm in Android environment (Android NDK, OpenCV, C++ & Java)
Week 6 - 12 April
- Implementation attempt of ORB tracking algorithm (C++ & OpenCV):
- Identify ROI and predict its position in next frame
- Verify in consecutive frames if ROI's objects matches with ORB
Week 30 March - 5 April
- Development continuation and study of a kernel-based tracking algorithm (C++ & OpenCV):
- Template matching
Week 23 - 29 March
- Optimization continuation of algorithms already implemented
- Development of a kernel-based tracking algorithm (C++ & OpenCV):
- Template matching using correlation (MOSSE Track)
Week 16 - 22 March
- Development of feature matching algorithms to detect a specific object (C++ & OpenCV):
- FAST + BRIEF
- Attempt of optimize actual and previous algorithms in terms of fps and accuracy
Week 9 - 15 March
- Optimization of video stabilization algorithms
- Implementation of one single object kalman filtering algorithm (C++ & OpenCV)
- Attempt of implement multiple objects kalman filtering algorithm
- Study of other algorithms based of feature matching and kernel-based tracking
Week 2 - 8 March
- Initialization of development of an algorithm of video stabilization using optical flow (C++ & OpenCV)
- Development of background subtraction using the video stabilization algorithm to detect objects moving faster than its background
2nd Meeting - 3rd of March
- Answer of the questions made in the last meeting
- Discussion of the work made until the moment
- Prof. Luís Teixeira gave some suggestions on how to approach the problem:
- Video stabilization using camera motion prediction
- Background subtraction for objects moving faster than the background
Week 23 - 28 February
- Continuation of mobile integration of HoG algorithm
- Analysis of mobile phone (Nexus 4) specifications
- Attempt of improving algorithm frames per second (fps) by optimizing code
Week 16 - 22 February
- State of the art study continuation
- Development of Histogram of Gradients (HoG) algorithm (C++ & OpenCV) to detect people in a video stream
- Initialization of mobile integration (Android NDK, OpenCV, C++ & Java)
Week 9 - 15 February
- State of the art study continuation
- Software installation:
- Eclipse Luna + CDT plugin
- Android SDK + ADT + NDK
- OpenCV
1st Meeting - 9th of February
- Discussion about the primordial aspects of the work to be made
- Algorithms able to provide the output desired
- Two questions made by Prof. Luís Teixeira:
- What was the main goal and purpose of this dissertation
- How and which methods could be used to perform an evaluation of the algorithm used
Week 2 - 8 February
- State of the art study initialization
- Research of tracking algorithms:
- Kernel-based tracking
- Visual feature matching
- Blob tracking - Research of filtering algorithms:
- Particle filter
- Kalman filter