OID_v4 Toolkit

• Toolkit by theAIGuysCode

Getting Started

Python is required.

1. Clone this respository

   git clone git clone https://github.com/EscVM/OIDv4_ToolKit.git

2. Install required packages with pip

   pip install -r requirements.txt

3. Download 1800 images of Person with the following command.

   python main.py downloader --classes Person --type_csv validation --limit 1800

   Arguments

   • --classes: Indicates the classes needed to be download
   • --type_cvs: Indicates the dataset type
   • --limit: Limit the number of dataset to be downloaded

   The algorithm will take care to download all the necessary files and build the directory structure like this:

   main_folder
   │   main.py
   │
   └───OID
       │   file011.txt
       │   file012.txt
       │
       └───csv_folder
       |    │   class-descriptions-boxable.csv
       |    │   validation-annotations-bbox.csv
       |
       └───Dataset
           |
           └─── train
                |
                └───Person
                     |
                     |000b65a36ad46f9e.jpg
                     |000e1dd786c8e433.jpg
                     |...
                     └───Label
                            |
                            |000b65a36ad46f9e.txt
                            |000e1dd786c8e433.txt
                            |...
   
   

For YOLOv4

1. Run convert_annotations.py. This will generate .txt annotation files in proper format for custom object detection with YOLOv4. The text files are generated in folder with images.

   python convert.annotations.py

For MobilenetSSD

1. Download the file and place in the OIDv4_toolKit parent directory

   Link: oid_to_pascal_voc_xml.py

2. Run the following command

   python oid_to_pascal_voc_xml.py

   The script will create directory called To_PASCAL_XML in the Dataset Subdirectories. These directories contain the XML files.