python voc parser
An easy way to read pascal voc dataset. For now it only has functionality for the detection competition.
Installation
Install from PyPi
pip install python_voc_parserInstall from source:
git clone https://github.com/gabrielrezzonico/python_voc_parser.git
cd python_voc_parser
pip install .Example usage
Extract the dataset, in this example the dataset was extrated in a folder called "VOCdevkit":
WORKING_DIRECTORY = os.getcwd()
# we are going to use all the train and validation images from all categories (trainval.txt)
IMAGE_SET_PATH = os.path.join(WORKING_DIRECTORY, 'VOCdevkit/VOC2012/ImageSets', 'Main')
DETECTION_COMPETITION_FILENAME = 'trainval.txt'
ALL_IMG_DETECTION_COMPETITION_FILEPATH = os.path.join(IMAGE_SET_PATH, DETECTION_COMPETITION_FILENAME)
# path to the voc annotation folder
ANNOTATIONS_PATH = os.path.join(WORKING_DIRECTORY, 'VOCdevkit/VOC2012/Annotations')
# Image path
IMAGE_PATH = os.path.join(WORKING_DIRECTORY, 'VOCdevkit/VOC2012/JPEGImages')#import the module
import python_voc_parser as voc
#create a parser object
parser = voc.VocAnnotationsParser(IMAGE_PATH, IMAGE_SET_PATH, ANNOTATIONS_PATH)# if you need a list with all the data:
annon_list = parser.annotation_line_list
# if you need a dataframe:
parser.get_annotation_dataframe()returns:
| class_name | height | img | img_full_path | width | xmax | xmin | ymax | ymin | |
|---|---|---|---|---|---|---|---|---|---|
| 0 | tvmonitor | 375.0 | 2008_000002 | /home/user/Personal/playground/voc/VOCdevkit/V... | 500.0 | 448 | 34 | 293 | 11 |
| 1 | train | 333.0 | 2008_000003 | /home/user/Personal/playground/voc/VOCdevkit/V... | 500.0 | 500 | 46 | 333 | 11 |
| 2 | person | 333.0 | 2008_000003 | /home/user/Personal/playground/voc/VOCdevkit/V... | 500.0 | 83 | 62 | 243 | 190 |
parser.get_annotation_dataframe_compact()returns:
| img_full_path | annon | |
|---|---|---|
| 0 | /home/user/Personal/playground/voc/VOCdevkit/V...2008_000002.jpg | (((34, 11, 448, 293), tvmonitor),) |
| 1 | /home/user/Personal/playground/voc/VOCdevkit/V...2008_000003.jpg | (((46, 11, 500, 333), train), ((62, 190, 83, 2... |
Contribute (Development of the package)
To develop a module based on this project follow this instructions.
Create a virtualenv for development packages
It's allways a good idea to create an enviroment for the module.
conda create --name packages_p3 python=3.5
source activate packages_p3
conda install virtualenv
virtualenv venvInstalling dev and test dependencies
pip install -e .[dev,test]Working in “Development Mode”
While you are developing the module you can install your module from source:
pip install -e .Test and coverage
To run the test:
$ py.test tests
==================================== test session starts =====================================
...
collected 4 items
tests/test_hello.py ...s
============================ 3 passed, 1 skipped in 0.02 seconds =============================To run the coverage:
$ py.test --cov=python_package tests/
==================================== test session starts =====================================
...
tests/test_hello.py ...s
----------- coverage: platform linux, python 3.5.3-final-0 -----------
Name Stmts Miss Cover
------------------------------------------------
python_package/__init__.py 1 0 100%
python_package/example.py 5 0 100%
------------------------------------------------
TOTAL 6 0 100%
============================ 3 passed, 1 skipped in 0.05 seconds =============================Convert README.md to README.rst
pandoc --columns=100 --output=README.rst --to rst README.md
