Skip to content

thepate94227/Keras-segmentation-deeplab-v3.1

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

80 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Keras DeepLab V3.1+

DeepLab V3+ for Semantic Image Segmentation With Subpixel Upsampling Layer Implementation in Keras

DeepLab is a state-of-art deep learning model for semantic image segmentation. Original DeepLabV3 can be reviewed here: DeepLab Paper with the original model implementation.

Features

  1. Conditional Random Fields (CRF) implementation as post-processing step to aquire better contour that is correlated with nearby pixels and their color. See here: Fully-Connected CRF
  2. Custom image generator for semantic segmentation with large augmentation capabilities.

New Features That Are Not Included In The Paper

  1. Keras Subpixel (Pixel-Shuffle layer) from: Keras-Subpixel for efficient upsampling and more accurate segmentation
  2. ICNR Initializer for subpixel layer (removing checkerboard artifact) ICNR
  3. Comparisson of the original Deeplab model with my Deeplab+subpixel+CRF
  4. Fast training - transfer learning from paper's proposed model to a better model within ~1 hour with 1-1080Ti GPU
  5. Jaccard (mIOU) monitoring during training process for multi-class segmentation tasks.
  6. Adaptive pixel weights.

Results

I've compared the segmentation visual results and the IOU score between paper's model and mine, as well as the outcome of applying CRF as a post-processing step.

Below depicted few examples:

alt text alt text alt text alt text

And the IOU score amid the classes:

alt text

I didn't receive a significant improvement of the IOU scores, perhaps due to low number of epochs. However I believe this method can eventually outperform the original model for a bigger dataset and more epochs.

Dependencies

  • Python 3.6
  • Keras>2.2.x
  • pydensecrf
  • tensorflow > 1.11

About

An awesome semantic segmentation model that runs in real time

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 82.2%
  • Python 17.8%