MajorProject-Text2Face

Title

Realistic Face Generation

Base Paper Implementation

StackGAN: Text to Photo-realistic Image Synthesis with Stacked Generative Adversarial Networks

Progressive Growing of GANs for Improved Quality, Stability, and Variation

Problem Statement

Although Conventional GANs have shown remarkable success in various tasks, they still face a lot of challenges in generating high quality images since generator and discriminator use the same backpropagation network. So, through this paper, we aim at generating realistic faces using a fully-trained GAN on different datasets.

Application Areas

• Movie Actors Casting
• Criminal face reconstruction
• Identification of victims for law agencies
• Image editing/Video games

Achitecture Diagram

Software/Hardware Requirements:

Software:

Language: Python Framework: PyTorch/Tensorflow Libraries: numpy, image_processing, scipy, matplotlib, pickle etc. Operating System: Windows7 (minimum)

Hardware:

RAM: 8GB (min.) Processor: core i5 or i7 HDD/SDD: SDD preferred (faster) Graphic Card GPU (might require later)

Team Members:

• Anukriti, DTU
• Anurag Mudgil, DTU
• Nakul Dodeja, DTU