trane293 / Object-Detection-Framework-Using-HOG-And-SVM

The project aims to provide a generic framework which utilized HOG as features and SVM as classifier to detect any objects that the user wants. The classifier can be trained to detect to detect anything. Just add positive images to pos folder and negative images to neg folder.

An Object Detection Framework Using HOG And SVM

##AUTHOR - Anmol Sharma

---

#Description of Files and Folders

    genfiles - FOLDER - Will contain SVM models that are trained on different training sets
	    specified by the naming scheme of different folders in which they 
	    will be stored. For example - 
	
	    svmlight_24   -  contains SVM model trained on 24x24 positive and
	                     negative images. This model is not trained using 
	                     hard negatives.

    svmlight_64   -  contains SVM model trained on 64x64 positive and 
        			       negative images. This model is not trained using 
        			       hard negatives.
    
    svmlight_24hn  - contained SVM model trained on 24x24 positive and
	         negative images. This model is also trained on a 
	         number of hard negatives obtained from a road video
	         and few from indoor scenes.

    svmlight_24hn_test - this folder will be used for storing test SVM models
		     which is not finalized and for prototyping purposes. 

    libsvm - FOLDER - should contain the libsvm.h header file. 

    nbproject - FOLDER - contains some important files that are used to configure the 
                        HOG training process. 

    pos - FOLDER - should contain positive images for training. 

    neg - FOLDER - should contain negative images for training. 

    sidefiles - FOLDER - contains some screenshots on how to set up this code. Useless.

    svmlight - FOLDER - contains important svmlight header files. These files were downloaded
				               from official website and then make build here. 

    test - FOLDER - should contain test data if required. 

    buildcode_main.sh - SHELL SCRIPT - Used to build the project using main.cpp as main file.

    buildcode_train.sh - SHELL SCRIPT - Used to build the project using main_train.cpp as main file.

    buildcode_prototype - SHELL SCRIPT - Used to build the project using main_prototype as main file.

    main_train.cpp - CPP FILE - Should contain the main training code to get training files, train an 
	                     SVM classifier and save the model to genfiles. 

    main.cpp - CPP FILE - Should contain the main code to test a pre-trained classifier ideally trained 
                        using main_train.cpp

    main_prototype.cpp - CPP FILE - Should contain prototype code in case user doesn't want to mess up 
		                  with main.cpp branch. 

---

###HOW TO BUILD CODE

Open terminal. Write: chmod +x <name_of_build_file>
example: chmod +x buildcode_main.sh
./buildcode_main.sh