Sem-CS: Semantic CLIPStyler for Text-Based Image Style Transfer (ICIP 2023 accepted, Conference Paper)

Semantic CLIPStyler Below steps to follow: Dependencies: This code works on Python 3.8>=

!pip install git+https://github.com/openai/CLIP.git

Other package dependencies are mentioned in requirements.txt

1) Salient Object Detection

We have provided already produced masks by Deep Spectral Segmentation here for now. We will update the code later for producing it in single pipeline.

2) Semantic CLIPStyler

a) To run single text condition
    python clipstyler_spectral.py --content_path "test_set/night1_resized.png" --segmentedImage_path "segmaps/night1_resized.npy" --filename "night1_resized" --exp_name "exp1" --text "Starry Night by Vincent van gogh"

b) To run multiple text condition
    python clipstyler_spectral_globfb.py --content_path test_set/night1_resized.png --segmentedImage_path segmaps/night1_resized.npy --filename night1_resized  --exp_name exp2 --textb "Pop Art" --textf  "Starry Night by Vincent Van Gogh"