KanikaGera / Self-Driving-Car-using-Deep-Learning

This repository has the implementation of a self-driving car's steering wheel component i.e. predicting the angle of rotation of steering wheel with respect to front board video as sensory input, using various deep learning algorithms.

Self Driving Car

Problem Statement

Problem is to implement a self driving car's steering wheel component with front board video as sensory input. The software component of deciding the angle of rotation of the steering wheel according to the curvature of the road is implemented in this project.Various deep learning methods are used and their performances is compared in this domain. Major Goal was be to reduce the training log loss and improve accuracy.

Structure of Repository

1. Folder 2DCNN & 3DCNN consist of jupyter notebook to train the model and model.py file that consist of model's architecture written with tensorflow api
2. Folder 2DCNN-LSTM consist of jupyter notebook to train model. Model's architecture is defined inside this notebook written with tensorflow keras api.
3. run.py is opnecv implemnetation to run the best model to predict the wheel's angle on given input sequence of images.

Dataset

Dataset is available on google drive link - https://drive.google.com/file/d/1GPp6Q2GpQcJSrsGpUUcRzLZpN7qjPaaX/view?usp=sharing

Getting Statrted- Requirements

1. Install tensorflow
2. Install opencv
3. Install jupyter notebook

Train Code

1. Open the folder according to the algorithm you want to train.
2. Open Notebook Present in it
3. Run the cells

Implementation Details

Detailed Report of Implementation and Analysis is avaliable in report at Project Report Folder.