readMe.txt

Read me

** NOTICE **
Although all ground truth training data are included in the zip file (can be used for detector re-training, more info in readMe.txt), the video sources for generating gTruth are too big to be uploaded. 
The sources are available at: 
https://drive.google.com/drive/folders/1wdLIo9EjiS7jbqhanAUZ-rbQaIvRUVbJ

1.
To run the code processing the video, please run the main.m file.
(Note that the first time running main.m could be slow, but after that the speed should be ok.)
It loads the data of detector files for different bird detection. 
Detection and tracking are developed throughout the processing of the Angry Bird video, with the function runBirdDetection.m
The location data of flying birds are recorded by the output of functions drawTraj.m and drawTraj_blue3.m.
After the video stops, the plots of bird trajectory are generated by running plot_bird.m.

2.
To re-generate the detector files that are loaded in the main.m file, please run the trainDetectors.m file.
It reads the Ground Truth data of different birds, and outputs the training detector files.

3. 
To figure out the process of bird labelling, please open 'Video Labeller' in the Apps Command in MATLAB.
By loading different video sources, labels can be assigned according to different algorithms. The output file is the Ground Truth data.
For more information, please watch the video at https://www.youtube.com/watch?v=ow_B_30WU1s

4. 
To obtain the panorama of red bird detection that uses a different approach to estimate the trajectory, please run the panoDetection.m file in the folder called pano.