Skip to content

Jayanths9/Plant_Image_Generation

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

10 Commits
 
 
 
 
 
 

Repository files navigation

Generating Synthetic Plant Images using GAN's

Introduction

Deep Convolutional Generative Adversarial Networks (DCGAN) are trained on plant images and used the trained model to generate synthetic plant images. In this project DCGAN is trained and evaluated on images from [64 x 63 x 3] upto [1536 x 156 x 3]. Along with StyleGAN3 is also implemented for image generation.

DataSet Preparation

Foreground segmentation using instance segmentation masks

original Image Insance Image Grayscale Image Segmeneted Image Final Cropped Image

(a) Original Image (b) Instance Segmentation of Original Image (c) Grayscale image of Instance segmentation image (d) Foreground Segmented Image (e) Final Cropped Image

Foreground Segmentation using Hsv color conversion

original Image Insance Image Grayscale Image Segmeneted Image

(a) Original Image (b)Image converted to HSV color scale from RBG (c) Extract plant mask based on color range (d) Bitwise multiply original image and plant mask

Generated Images from DCGAN

Generated Images for 1536 X 1536 dimension

1536 Image 1536 Image

Generated Images for 1024 X 1024 dimension

1024 Image 1024 Image

Generated Images for 256 X 256 dimension

256 Image 256 Image

Generated Images for 64 X 64 dimension

64 Image 64 Image

Generated Images from StyleGAN3

256 X 256 Image dimensions

256 X 256 Images 256 X 256 Images 256 X 256 Images 256 X 256 Images

512 X 512 Image dimensions

256 X 256 Images 256 X 256 Images 256 X 256 Images 256 X 256 Images

Model Architecture

Generator and Discriminator Architecture

GAN architecture

Author

Jayanth Somashekaraiah
Universität Bremen, Bremen
Email: jayanth@uni-bremen.de

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published