The project aims at taking in a Content Image and asks the user to select a Style Image and uses certain layers of VGG-16 which is a pre-trained model to merge the content and the style image to give an Output / Generated Image.
- Aastha Chowdhary
- Manan Poddar
- Anshuman
- Ankush
- Rushad
- We created an Android app that gives the user the option to select the style image from the various style images that we offer and also select any content image from their camera roll. On clicking the button an Output image is generated which is a combination of the 2 images selected. This output image is such that it has retained the content of the content image and the style of the style image.
- For the backend, we have started off with a white noise image and used certain layers of VGG-16 some of which are known to retain the content and the others which are used to retain the style. We have tried to minimize the loss for both the content and the style to train the white noise image such that it retains both the above and has a minimum loss.
- For the backend, we have used Django and for the frontend, we have used React Native.
- The whole program is hosted on our laptop and we have used Expo to connect our mobile to the laptop.
8 weeks. 2-3 hours a week.