Overview / Usage

It is a mapping of unpaired images, created by extracting features of one image and transforming and imposing them on another image. A one-to-one mapping between images from input to target domain in the training set is very powerful and can be used in variety of problems for solving real life problems. This is achieved by a type of generative model, specifically a Generative Adversarial Network dubbed CycleGAN.
Here we map the color pallet to the extracted wall feature of the room Image. Using CycleGAN and simple CNN model to extract the latent space of the wall feature from the room Images and then mapping it to the color pallets.
In today world we face many problem and one of them is money we waste on the average interior designing. On average interior designing costs from $500- $5000. Which is a huge problem for a common man to get. So, we wanted to do something for them so they can get the professional advises at minimal cost or no cost without wondering about the money. In market there are many company provide some kind of color choosing mechanism which are not good enough to give satisfaction to the customer where you have to work and the result is not that good.

Methodology / Approach

First we took the different images of the Room where we can extract the features of wall and edge from the room images using simple CNN and encoder-decoder program. Then after that we used the CycleGAN for the mapping of the features of the wall to the pallets in process we find out that the when pallet features are mapped on the wall featyures we get better results. For better result we have to increase the epoch because after every 100-200 epoch it was able to distuiguish the edges and corners.
In process we used Tensorflow, Numpy, GANs, etc

Technologies Used

libraries used are Tensorflow, Numpy, Pandas, Scipy, Pilllow
Software used is linux image conversion tool
Hardware used are GTX1080
Technologies used is Floydhub, Tensorboard

Repository

https://github.com/ksajan/AI-Interior