Skip to content

This repository implements a neural network-based model for road segmentation of images from the KITTI Road dataset using an approach based on the U-NET (encoder-decoder) architecture. In this Jupyter Notebook code, we have all the processes, from model creation to dataset splitting, training, validation, and testing on images and videos.

Notifications You must be signed in to change notification settings

Vinicius-GN/VGG16_RoadSegmentation

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 

Repository files navigation

Road Segmentation using VGG-16 and U-Net Architecture

Overview

This repository implements a neural network-based model for road segmentation of images from the KITTI Road dataset using an approach based on the U-NET (encoder-decoder) architecture. In this Jupyter Notebook code, we have all the processes, from model creation to dataset splitting, training, validation, and testing on images and videos.

VGG-16

VGG-16 is a deep convolutional neural network originally designed for image classification tasks. In this implementation, we use VGG-16 as the encoding part (encoder) of our U-Net architecture to leverage VGG-16's feature extraction capabilities.

VGG

U-Net Architecture

The U-Net architecture consists of two main parts: an encoder and a decoder. The encoder applies a series of convolutions and pooling to capture features from the image, while the decoder applies convolutions and upsampling to reconstruct the segmented image.

VGG

Results

To facilitate training with limited computational resources, we used 128x128 pixel images. Below are examples of segmentation results:

Result Examples

1 2

4+ 4+

Requirements

  • TensorFlow
  • OpenCV
  • Matplotlib
  • Tqdm

Contributions

Contributions are welcome! Feel free to open issues or pull requests.

About

This repository implements a neural network-based model for road segmentation of images from the KITTI Road dataset using an approach based on the U-NET (encoder-decoder) architecture. In this Jupyter Notebook code, we have all the processes, from model creation to dataset splitting, training, validation, and testing on images and videos.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published