Getting Started

The yoloxm.onnx is the onnx file generated from yolox training.

You can only activate the calculation feature without onnx model.

The onnx file use in this application is trained on 640x640 images with yolox_m model. [Yolox Github]

Python script deployment

You can simply execute the python file, after you installed all the requirement packages.

The python version is 3.8.3

git clone https://github.com/ZachKLYeh/Yolox_Annotation_Application.git

cd Yolox_Annotation_Application

pip3 install -r requirements.txt

python main.py

Pyinstaller deployment

Also, this script can be wrap up into executable file via pyinstaller.

You can generate the executable application by executing these commands.

The executable file reside in Yolox_Annotation_Application/dist/main.exe

git clone https://github.com/ZachKLYeh/Yolox_Annotation_Application.git

cd Yolox_Annotation_Application

pip install pyinstaller

pyinstaller main.py --onefile --add-data "yoloxm.onnx;." --windowed

cd dist

main.exe

Feature Introduction

Main Window

Generate xml format annotations:

• Select the input folder and output folder

Input folder can only contain jpg images, other files will raise an error. The default setting is to set the output folder the same as the input folder. You can also use the button "..." to utilize file browser.

• Select threshold

The threshold is the confidence threshold of prediction, higher threshold will result in higher accuracy. However, the detected object will be less. The default threshold is 0.5(maximum = 1.0).

• Select whether to visualize

If you select the visualize option, this will create a "visualized_images" folder in your designated output folder. However, if there's already a "visualized_images" folder, this will raise a warning. Making sure you are notified of overwriting these images.

• Press "Generate annotations" button

This will generate the result, both annotations and visualized images. After the generation. There will be a pop up window, indicating the annotation path and visualizing images path.

Caluculate annonation instance:

• Select input folder

The calculation only takes the input folder. Other options are not considered.

• Press "Calculate annotations" button

This will calculate label instances in the input folder, also sub folders in the input folder. The result will be reported in a pop up window after calculation.

Warning Messages

Overwrite warning

This is the case when your designated output folder already have labeled data or visualized images. There will be a warning, you can decide whether to overwrite the files.

Error Messages

File format error

If the input folder exist some files except for jpg and xml format, this will raise an error.

Invalid path error

If your input folder or output folder path does not exist, this will raise an error.

Invalid label error

If your xml file have bonding box location that exceed image size, this will raise an error. The error will report the xml file that has invalid bonding box location.