Yolo Annotation Tool-New

Manivannan Murugavel
2 min readMar 5, 2019

--

This blog explain how to use new yolo annotaion tool.

This is Yolo new annotation tool for annotate the image for yolo training. I have posted three blogs for how to train yolo with our custom objects or images.

In this three blogs use three steps for annotate the images.

Old Steps:

run main.py
run convert.py
run process.py

The main.py used for draw the bounding box on image and stored the top-left and bottom-right points in corresponding txt file. Example txt file below

2
297 128 367 316
352 116 388 228

The convert.py used for convert the above points into yolo input format. The yolo input format is

Syntax:

class_id x y width height 

Example:

0 0.51875 0.616666666667 0.109375 0.522222222222
0 0.578125 0.477777777778 0.05625 0.311111111111

Note: x and y is not top-left, it is center point of the bounding box.

The process.py is used for split the train and test dataset for yolo training.

New Steps:

Now we are only use two steps for yolo training.

run main.py
run process.py

I have implemented the new feature in main.py. This python script generated the yolo input format txt file for corresponding image file.

Please see the below demo video for how to generate the txt file.

Full source code is here, Please clone it and run the code.

I have tested with my new annotation tool. If you have any bug or need any options please raise the bug from github.

--

--