tapos-datta / imagestyletransfer

A project that can compose an image into a cartoon or the style of another image by using pre-trained deep learning models.

Image Style Transfer

A project that can compose an image into a cartoon or the style of another image by using pre-trained deep learning models.

A GLSurfaceView is used for rendering image into view and the image data is transfered from view to ML models, and vice versa, through a pixel buffer rather than bitmap. This GLSurfaceView provides a OpenGL ES context which will be effective in further pre/post processing.

Used Deep Learning Models

• CartoonGAN model
• Arbitrary-image-stylization

Image Resources

• Image resources as Styled Images or Sample Input are picked from Unsplash

Outputs

Acknowledgments

• Thank you to @MasayukiSuda for his work on Mp4Composer-android project where OpenGL ES components are organized very well.
• Thanks to Google TensorFlow team and the people who review and publish TFLite version of ML models.