/SEMANTIC-IMAGE-SEGMENTATION

This repository implements and evaluates SegNet, U-Net, and models from segmentation_models.pytorch for vehicle segmentation using PyTorch. It showcases these models' capabilities in accurately segmenting vehicles in complex scenes, demonstrating cutting-edge deep learning techniques in computer vision

Primary LanguageJupyter Notebook

Car Segmentation Dataset and Models

This repository contains implementations of various semantic segmentation models applied to a car segmentation dataset. The dataset comprises images of cars with their segmentation masks, allowing for the evaluation of these models' ability to accurately segment cars in images.

Features

  • Dataset: Contains images of cars with their segmentation masks.
  • Models: Implements SegNet, U-Net, and a pre-trained model using segmentation_models.pytorch.
  • Evaluation: Includes metrics for evaluating model performance, such as pixel accuracy and mean IoU.

Getting Started

Prerequisites

  • Python 3.x
  • PyTorch
  • segmentation-models-pytorch

Installation

Clone the repository and open in google colab.

Data Preparation

Download the dataset from the provided link and place it in the appropriate directories (images and masks)

Model Implementations

SegNet

  • Description of SegNet architecture and its relevance to semantic segmentation.

U-Net

  • Description of U-Net architecture and its relevance to semantic segmentation.

Pre-trained Model

  • Description of the pre-trained model using segmentation_models.pytorch and its benefits.

Usage

Include usage instructions and code snippets for loading models, preparing data, and performing predictions.

Evaluation

Explain the evaluation metrics used and how to calculate them using the provided functions.