Skip to content

liuch37/semantic-segmentation

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

35 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Semantic Segmentation

Simple inference implementation with trained HRNet on MIT ADE20K dataset, using PyTorch 1.6.0. Most of the code taken from [1]. Purpose of this project is to unify sky pixels with ultra high prediction confidence to a single color, in order to remove sun light effect and inconsistent cloud texture.

Usage

  1. Download pretrained model from http://sceneparsing.csail.mit.edu/model/pytorch and store them in ./ade20k-hrnetv2-c1/.

  2. Specify your test input image directory and test output image directory in semantic_segmentation.py.

   image_path = './input/'

   output_path = './output/'
  1. Specify if you want to use GPU or not (-1 for cpu, others for gpu index) in semantic_segmentation.py.
   gpu = -1
  1. Run python semantic_segmentation.py

  2. All the inference results will be stored in your output_path

Results

semantic map

Image of semantic map

sky filtered image

Image of semantic map

Source

[1] Original code: https://github.com/CSAILVision/semantic-segmentation-pytorch.

[2] HRNet: https://arxiv.org/abs/1904.04514.

About

Simple inference implementation with trained HRNet.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages