Oneshot Character Extraction From Anime Video With MultiModal Method

This is demo code in order to extract character from anime video with multimodal method under a oneshot setting.

With these multimodal prompt:

• text prompt ( describe the character with some words,like one anime girl with long black hair and blue eyes)
• image prompt ( one image of the character, selected from the anime video is best )
• audio prompt ( the audio segments of the character) NOT IMPLEMENTED FOR WITH THE ABOVE TWO METHOD THE RESULTS ARE GREAT ENOUGH

Besides the above multimodal prompt, I also add a anime face detector module from this project and it improves much.

• LAVIS - A Library for Language-Vision Intelligence
• Anime Face Detector