Generate_Dataset : Pipepeline in order to generate a dataset for Fast-RCNN_inception_v2 model from green chromakey(Python version)

This project was created with the help of Jiri Sedlar. This repository host the back-end module for the project iReal. For a general description of the project, please refer to the iReal repository

Descripton of the project

Mandatory folders

The main script used is generate_dataset.py. It must be near folders named:

• mask_augmented Mask png created from the dataset generation

• rgb Green chromakey pictures where there is objects. Here it is inria remote and inria phone (70 pictures each)

• rgba_augmented Final PNG from the dataset generation with rgb mixed with bg

• xml Xml template used for the dataset generation

• xml_augmented Where the xml in correspondance of rgba_augmented are created. It must be used with rgba_augmented in order to create a Fast-RCNN_v2_model

Additional folder and script

• Dataset_7000 The generation of a dataset with 7000 pictures (3500 for both objects phone and remote)

• modify_name_in_xml.py If you want to modify informations inside the xml

Launch the dataset generation

python generate_dataset.py