Skip to content

Automatically remove background image/text/noise from images with foreground object

Notifications You must be signed in to change notification settings

sandeepchittilla/carvana-image-segmentation

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 

Repository files navigation

carvana-segmentation

Automatically remove background image/text/noise from images with foreground object

Dataset

We use the car dataset from the Kaggle competition : Carvana Image Masking Challenge for this purpose.

Solution

We model this as an Image Segmentation problem and build a Convolutional Neural Network (CNN) for classification between foreground and background. The dataset has ~5K images of cars with masks. We use the U-Net architecture to build the CNN and train it on ~4K images and test on ~1K.

The Python notebook carvana_segmentation.ipynb has code for :

  • Preliminary EDA
  • Image Preprocessing (Resize, Reshape, Image Augmentation etc.)
  • Defining a CNN with U-Net Architecture
  • Training, Model Validation and Quality Assesment Metrics (IoU, F1, PR curves)
  • Predictions on images outside Carvana Dataset

There is also a .py notebook to easily identify changes between commits

About

Automatically remove background image/text/noise from images with foreground object

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published