2vin / yolo-annotation

YOLO-style annotation tool for preparing dataset of a new object

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

yolo-annotation

YOLO-style annotation tool for preparing dataset of a new object from videos directly

This program can be used to prepare/ annotate new class in YOLO

Getting started

  1. Record videos of the object to be annotated. For instance, we will annotate oranges fromt the sample video.
  2. Make an empty directory to save annotation results. For instance, dataset in this case.
  3. Compile the program using following command:
    g++ annotate.cpp -o main `pkg-config --cflags --libs opencv`
  4. After successfully compiling the program, run the following command to start annotation process.
    ./main video_path output_folder
    In this case, ./main ./oranges.mp4 ./dataset/
  5. This will open a frame window where you can select the objects using mouse.
  6. Once all the objects of defined class are selected in the image, double press 's' to confirm the selections to be saved.
  7. This will generate the YOLO-style annotation files for the object from the video.

Program in action

Alt Text

The annotated files will be saved like this in the output folder
Alt Text

About

YOLO-style annotation tool for preparing dataset of a new object

License:GNU General Public License v3.0


Languages

Language:C++ 100.0%