GSOC 2015 project for Red Hen Labs. More details here http://vasanthkalingeri.github.io/CommercialDetection/
##Instructions to run
*Run dependencies.sh script
*Create a database in Mysql to store fingerprints
*Open src/constants.py, edit the line
CONFIG = {
"database": {
"host": "127.0.0.1",
"user": <username>,
"passwd": <password>,
"db": <name of the database created>
}
}
###To teach new commercials
*Create a file like data/labels for the given video
python src/main.py -g data/video.mpg data/labels
*The above creates a directory called db containing the commercials labelled. The system also learns from these commercials
###To detect commercials
python src/main.py -r data/test.mpg
*This will create a file called output.txt containing the detected commercials.
*It will have unclassified sections which can later be seen and edited. The edited file can be used as labels file for learning new commercials.
*Edit content in output.txt after watching the video manually.
*Now using output.txt as labels for test.mpg, run the program again to update the db with new commercials.
python src/main.py -g data/test.mpg output.txt
##Results The program was tested on Ubuntu 14.04 64 bit LTS. With the following versions of libraries
*Opencv cv = 2.4.8
*Numpy = 1.8.2
*Scipy = 0.13.3
*ffmpeg = 2.6.2