Semantic Segmentation Networks Keras Implementation

Brief Description

These are implementations for some neural network architectures used for semantic segmentation using the deep learning framework "Keras".

The Network Architectures Included:

1. Unet Architecture: https://arxiv.org/abs/1505.04597
2. Modified Unet with Resnet and VGG as Encoder.
3. Segnet Architecture (Resnet Encoder): https://arxiv.org/abs/1511.00561
4. Modified Segnet with Resnet and VGG as Encoder.
5. DeepLabv3: https://arxiv.org/abs/1706.05587

Utilized Technologies and Frameworks

• Keras (Tensorflow Backend).
• Basic Python Data Manipulation Packages (Matplotlib,....etc).
• SciKit Learn.

Repository Structure

1) Model Files:

• 5 files for the implementations of the previously mentioned architectures.

2) Utilities Files:

• util_funcs.py: contains some helper functions and the metric function.
• preprocessing.py: contains the code for preprocessing data before training.

3) train.py:

• contains training setup for differect models.

3) test.py:

• contains test setup.

How to Use it

• First, Clone the repository.

• To run model training:
  -- First, you need to edit "train.py" file to set some variables indicated inside.
  -- Then, run "train.py".

• To run inference:
  -- First, you need to edit "test.py" file to set some variables indicated inside.
  -- Then, run "test.py".

  Thanks a lot ^_^