[6] - Paper by Kolekar et. al(references in report.pdf)


	
*****************************************************Files**************************************************************/
--ai.sql
	-database of all results obtained by us.
************************************************************************************************************************
**Disclaimer** - All the code has been written by us(even for [6]), *except* requirements in Level-1.

Level-1
Audio analysis:
	- run ste_zcr.m
		- input -> an audio or video file(pass input in first line of this file).
		- output -> All excitation frames.
		- requirements -> energy.m, sgn.m, winconv.m, zerocross.m
	
Level-2 : Replay detection
	- run replays.m
		- input -> a video(preferably .mp4, otherwise may throw error for plugins for .avi, .mkv).
		- output -> 'images' folder containing all frames of the video, named img%d.jpg.
		- requirements -> no other files, match video.
	- Approach by [6]
		- run histo_p.m
			- input -> 'images' folder <- created above, 'replay.jpg' <- reference replay frame. check frames range in histo_p.m, all those frames must be in the folder.
			- output -> All matched frames corresponding to replay.
	- Our Approach
		1. Replay frames using correlation coefficient.
		- run replay_detection.m
			- input -> 'images' folder <- created above, 'replay.jpg' <- reference replay frame. check frames range all those frames must be in the folder.
			- output -> All matched frames corresponding to replays.
		- run scorebar_p.m
			- input -> 'images' folder <- created above, 'scorebar.jpg'(im1 in line 3) <- reference scorebar, 'any other image for frame size'(im2 in line 4). check frames range, all those frames must be in the folder.
			- output -> 'output2.csv', contains [img name, HHD(difference), yes/no], yes if it doesn't have scorebar.
			
Level-3 : Field view detection
	- Approach by [6]
		- dgpr_p.m
			- input -> 'images' folder <- created above, check frames range, all those frames must be in the folder.
			- output -> 'output.csv', contains [img name, DGPR value, yes/no], yes if belongs to field view.
	- Our approach
		- fieldview_svm.m
			- the test set and training set are in array [x, y, k], where x is start frame, y is end frame, and k is the class.
				Thus if [3204, 6789, 1] is in any set, this means frames from 3204 to 6789 belong to class 1.
			- input -> 'images' folder <- training set and test set, created above, check frames range, all those frames must be in the folder.
			- output -> 'fieldview_svm.csv', contains [img name, class by svm, class manual], yes if belongs to field view.
				- if class manual is 'manual', -> this is a training frame.
		
Level-4a : Field View Classification
	-Approach by [6]
		- fieldview.m
			- input -> 'images' folder <- created above, check frames range, all those frames must be in the folder.
			- output -> 'fieldview_kolekar.csv', contains [img,FP,FP11,FP12,FP2,class by matlab,class by matlab]. yes class by matlab is 	printed twice(for mysql table).
	- Our approach
		- pitch_svm.m
			- the test set and training set are in array [x, y, k], where x is start frame, y is end frame, and k is the class.
				Thus if [3204, 6789, 1] is in any set, this means frames from 3204 to 6789 belong to class 1.
			- input -> 'images' folder <- training set and test set, created above, check frames range, all those frames must be in the folder.
			- output -> 'pitchview_svm.csv', contains [img name, class by svm, class manual].
				- if class manual is 'manual', -> this is a training frame.
		- long_svm.m
			- the test set and training set are in array [x, y, k], where x is start frame, y is end frame, and k is the class.
				Thus if [3204, 6789, 1] is in any set, this means frames from 3204 to 6789 belong to class 1.
			- input -> 'images' folder <- training set and test set, created above, check frames range, all those frames must be in the folder.
			- output -> 'longview_svm.csv', contains [img name, class by svm, class manual].
				- if class manual is 'manual', -> this is a training frame.
		- boundary_svm.m
			- the test set and training set are in array [x, y, k], where x is start frame, y is end frame, and k is the class.
				Thus if [3204, 6789, 1] is in any set, this means frames from 3204 to 6789 belong to class 1.
			- input -> 'images' folder <- training set and test set, created above, check frames range, all those frames must be in the folder.
			- output -> 'boundaryview_svm.csv', contains [img name, class by svm, class manual].
				- if class manual is 'manual', -> this is a training frame.

Level-4b : Non-Field view classification
	- Approach by [6]
		- close_up.m
			- input -> 'images' folder <- created above, check frames range, all those frames must be in the folder.
			- output -> 'output1.csv', contains [img, EP value,class by matlab,class by matlab].
	-Our approach
		- close_crowd_svm.m
			- the test set and training set are in array [x, y, k], where x is start frame, y is end frame, and k is the class.
				Thus if [3204, 6789, 1] is in any set, this means frames from 3204 to 6789 belong to class 1.
			- input -> 'images' folder <- training set and test set, created above, check frames range, all those frames must be in the folder.
			- output -> 'closeup_crowd_svm.csv', contains [img name, class by svm, class manual].
				- if class manual is 'manual', -> this is a training frame.

Level-5 : Crowd classification
	- Our approach
		- close_classify.m
			- the test set and training set are in array [x, y, k], where x is start frame, y is end frame, and k is the class.
				Thus if [3204, 6789, 1] is in any set, this means frames from 3204 to 6789 belong to class 1.
			- input -> 'images' folder <- training set and test set, created above, check frames range, all those frames must be in the folder.
			- output -> 'fielder_crowd_svm.csv', contains [img name, class by svm, class manual].
				- if class manual is 'manual', -> this is a training frame.
	
*********************************************************************************************************************************************