Settings

General settings

Balloon Detection

The tool’s balloon (bubble) detection uses offline balloon detection models trained by Darknet Yolo or TensorFlow Object Detection API.

Existing offline model: balloon detection model for manga.

It can also uses object detection services provided by Baidu EasyDL and Azure. You need to create your own account and train your own model.

Recommended labeling methods:

Box text areas in all balloons. Let the box fits the text as closely as possible.

_images/image_annotation.JPG

After the model is published, you can set the URL through File-Preferences-API setting. The corresponding API key is also required.

Text Image Recognition

By default, the software uses the local model to recognize text and non text areas. You can also call an API service.

The default address is: http://127.0.0.1:8082/classify

Please run the code in this repository: https://github.com/xulihang/text-image-classifier.

Natural Scene Text Detection

EAST and CRAFT natural scene text detection methods are supported.

The default address is: http://127.0.0.1:8080/detect

Code repository: https://github.com/xulihang/ImageTrans_SceneText_Detection.

API Settings

Click File->Preferences->API to set up APIs for OCR and machine translation services provided by Baidu, Youdao, Microsoft, Tencent, etc.

_images/api_setting.jpg

The following is a list of services that need to set the API key.

OCR:

For Machine translation, see BasicCAT’s docs.

Appearance Settings

Go to File->Preferences->Theme to adjust the appearance. In addition to the default theme, there are black and green themes.

You can also use external CSS to adjust the appearance of the software.

For example, the following CSS can control the text size of the text editing areas:

.text-area {
    -fx-font-size: 25 !important;
}

The font of the text editing areas can also be configured in project settings.