Overview / Usage

We used CycleGAN that basically do the mapping between the two given object. And here objects are features which are extracted from the two given pictures. In this project we tried to manipulate the one function where in one function the only feature which should it extract is wall by training it first by to detect wall and extract feature of wall from some pictures. Then this extracted feature from the picture are mapped to the available(already given) colour pallets. In this project we tried to solve the hectic job and very costly job of just designing job we made easy and cheap. This project can many people who have low cost budget and want to design their home. without going to some expert and asking their which colour will best and all that stuff. Their are many products available for this kind of stuff but all of them are bad at user experience and very inconvenient to use so we made this project which can be used as commercial product if this project will complete.

Methodology / Approach

First we disintegrated the code behind CycleGAN code and then tried to add some feature so it can used for this project. After Learning how CycleGAN works we integrated it with simple CNN where we tried to extract and train one model which will only try to extract wall feature from picture given from user after this we integrated with CycleGAN. Then we mapped these extracted feature from the given picture of house by user with the available colour pallets. In this project We use Tensorflow Framework and used GAN(Generative Adversarial Networks), CNN(Convolutional Neural Network) for training and extracting the feature from the pictures provided by the user.

Technologies Used

Libraries:- NumPy, Matlab, Tesorflow, Glob, Time, os, SciPy
Software:- Microsoft Visual Studio Code, Anaconda, Google Picasso, Ubuntu Terminal, Tesorflow, TensorBoard, Nvidia,
Hardware:- Nvidia GTX1080 Workstation with 250 GB SSD storage, 16GB RAM