tactile-texture-transfer

Transferring real and tactile textures made by users onto digital images using neural style transfer.

Requirements

Tensorflow

Keras

OpenCV

Numpy

Matplotlib

Usage

1. Clone Repository git clone https://github.com/parthnatekar/tactile-texture-transfer.git and navigate to TTT/

2. Install the IP Webcam app from the Google Play Store on your mobile device.

3. Start a webcam server on the application.

4. Run python3 ttt.py -p PATH_TO_INPUT_IMAGE -i IP_ADDRESS_OF_MOBILE_SERVER:PORT

5. Use flag -m (optional, default static) for static (i.e the style image is captured only once before the transfer algorithm is run) or dynamic (the style image is captured every epoch while running the algorithm, allowing modification of target style during transfer). Eg: python3 ttt.py -p PATH_TO_INPUT_IMAGE -i IP_ADDRESS_OF_MOBILE_SERVER:PORT -m dynamic

Results

The image below shows an example implementation, where a user creates a texture using grains, and transfers this to an input image.