DuNGEOnmassster/Multimodal_Style_Transfer

Image Style Transfer with style comes from text description(CPU friendly).

Multimodal_Style_Transfer

Image Style Transfer with style comes from text description(CPU friendly).

TODO ideas

(done) Style comes from text description

() Style comes from anothor image(classic style transfer)

() Style comes from audio description

() Real-time Transfer

(done) Expand text description with ChatGPT API

Setup

You are recommended to create a virtual envirnment with conda

conda create --name StyleTransfer --file requirement.txt

If failed, You can also setup virtual envirnment step by step

conda create -n StyleTransfer python=3.10

conda activate StyleTransfer

Then download the necessary Python packages through

pip install -r requirement.txt

Run

Run StyleTransfer with

python transfer.py --content_path <content_img_file_path> --text <style_description> --output_path <target_img_storage_path> --exp_name <exp_name>

Of course you can just simply start and adjust parameters in code

python transfer.py

And here are examples for what it would look like if you use Van_Gogh_Horus as text description.

Content	Target

Hopefully you will like it.

Have some fun

After obtaining the authorization from 'victim', we generated the following image with alias from JOJO.

Content	Crazy diamond	Golden experience	Star Platinum

Reference

Gatys et al., 2016 Gatys, L. A., Ecker, A. S., & Bethge, M. (2016). Image style transfer using convolutional neural networks. Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 2414–2423).

DIVE INTO DEEP LEARNING (14.12. Neural Style Transfer)

CLIPstyler: Image Style Transfer with a Single Text Condition

StyleCLIP: Text-Driven Manipulation of StyleGAN Imagery