Informations

With this image segmentation model, you can easily detect speech bubbles in comics or manga and obtain a mask to perform actions on them.

Note that the mask segmentation is often wavy at the edges of the detected speech bubbles.

Models

This repo contains the original .pt model and a .onnx model expored with the dynamic argument set on true.

Try out my model with your own pictures now by using my huggingface space here.

Previews

[!NOTE] If you need a model to detect speech bubbles, you might be interested in my attempt at automating the colorization of 2D pictures/drawings/comics using SD models. Learn more here.