VOC_Annotation

This is a VOC-like Database Annotation Program.

Introduction

VOC_Annotation is a Matlab-based Program which is used to make annotations for PASCAL VOC-like database, such as the database built by ourselves.

This program is designed and written by 5HU W@N9 in Dec 2015.

License

VOC_Annotation is released under the GPL v2 License (refer to the LICENSE file for details).

Contents

1. File Structure
2. Requirements
3. Program List
4. Usage

File_Structure

1. JPEGImages - This folder is used to store the original images which are saved as JPEG format.
2. Annotations - This folder is used to store the annotation files of original images which are saved as .xml format.
3. GTImages - This folder is used to store the ground truth images that make the annotation files visiable.
4. ImageSet - This folder is used to store the image list file specifying which images are used in the training process.
5. Segmentation - This folder is used to store the segmentation images which are segmented from the original images.
6. SelectiveSearchData - This folder is used to store the selective search results of original images.
7. Program - This folder is used to store all program files, and you can see the Program List for more details.

Note The Database only contains JPEGImages/, Annotations/, ImageSet/ and SelectiveSearchData/.

Requirements

1. Matlab R2014a or higher version
2. g++ (optional for part of programs)
3. OpenCV 3.0+ (optional for part of programs)
4. gtk-config (optional for part of programs)

Program_List

1. preproc.cpp - used to convert the original video into the dividual video frames which are saved in _cache/.
2. preproc.m - used to pre-process the video frames into the original images saved in JPEGImages/ such as cut, resize.
3. demo.m - used to make annotations for original images, generating the .xml files, GT images and Segmentation images.
4. makelist.m - aim to make a train image list which is saved in ImageSet/Main/.
5. demo_proposals.m - used to make proposals for original images which will saved in SelectiveSearchData/.

Usage

1. preproc.cpp

Type the following command to compile preproc.cpp :

$ sudo chmod a+x compile.sh
$ ./compile.sh 

This command will make the folder _cache/ and generate the executable file preproc.

Execute the preproc program to convert the video into dividual video frames

$ ./preproc PATH_TO_THE_VIDEO

2. preproc.m

Open Matlab R2014a and run this file.

$ matlab -nodesktop -nosplash
>> preproc

Once you finish running this program, you can delete the _cache folder safely.

$ rm -r _cache

3. demo.m

This part of illustration is written in Program/README.txt which is written by Chinese language.

You can check this file and see what you should do in details.

4. makelist.m

Open Matlab R2014a and run this file.

$ matlab -nodesktop -nosplash
>> makelist

And you can see the imagelist.txt in ImageSet/Main/.

5. demo_proposals.m

Open Matlab R2014a and run this file.

$ matlab -nodesktop -nosplash
>> demo_proposals

And you can see the car_2015_train.mat in SelectiveSearchData/.