This project focuses on performing multiclass segmentation on images using two different neural network architectures: U-Net and a custom-designed segmentation network. The goal is to accurately segment objects in images into various classes.
Notebook deploy model: https://colab.research.google.com/drive/12B9kXZbK_cSuO9HXucJu7QglPavnEGKy?usp=sharing
Program language: Python
Framework: tensorflow
Model architecture:
In this project, I explore two different neural network architectures for multiclass segmentation:
-
U-Net: consists of an encoder-decoder structure that captures both local and global features for accurate segmentation.
-
Custom Segmentation Network: Apart from U-Net, I design and implement a custom neural network architecture tailored to the specific multiclass segmentation task.
pip install -r requirements.txt
To train the model, execute the following command:
python /path/to/train.py --config "/path/to/config/file"
To perform testing, use the following command:
python /path/to/test.py --config "/path/to/config/file"
The image on the left hand side is the ground truth and the right hand side is the prediction
