Project information
- ๐ฏ Objective:: The goal of the project was to train a Deep Network to replicate the human steering behavior while driving, thus being able to drive autonomously on a simulator provided by Udacity. To this purpose, the network takes as input the frame of the frontal camera (say, a roof-mounted camera) and predicts the steering direction at each instant.
- date: Sep 2022 - Oct 2022
- Github Link: Github Link
Description:
๐ก Features:
- ๐๐๐๐ฉ ๐๐๐๐ซ๐ง๐ข๐ง๐ ๐
๐ซ๐๐ฆ๐๐ฐ๐จ๐ซ๐ค: Utilizes Keras with Convolutional Neural Networks (CNNs), leveraging deep learning to process and interpret visual data for steering control.
- ๐๐ข๐ฌ๐ฎ๐๐ฅ ๐๐ง๐ฉ๐ฎ๐ญ ๐๐๐ง๐๐ฅ๐ข๐ง๐ : The network takes input from three different camera perspectivesโfrontal, left, and rightโmounted on the vehicle, allowing it to understand and respond to varying road conditions and angles.
- ๐๐๐๐ฅ-๐๐ข๐ฆ๐ ๐๐ญ๐๐๐ซ๐ข๐ง๐ ๐๐ซ๐๐๐ข๐๐ญ๐ข๐จ๐ง:
-Direct Control: Predicts steering directions frame by frame based on the visual input, enabling dynamic and responsive vehicle control.
-Behavioral Cloning: The network learns directly from human driving behavior, effectively cloning the steering actions based on visual cues.
Python and Keras: Built with Python and utilizing the Keras library for neural network construction, facilitating an efficient and powerful computational framework suited for real-time applications.
- ๐๐ซ๐๐ข๐ง๐ข๐ง๐ ๐๐๐ญ๐ ๐๐๐ช๐ฎ๐ข๐ฌ๐ข๐ญ๐ข๐จ๐ง:
-Simulator Training Mode: Collects data in a controlled environment where a human driver navigates the simulator, capturing both the visual frames and the corresponding steering directions.
-Pre-compiled Dataset: Utilizes Udacityโs pre-built dataset which includes 8036 samples with images from three different camera views and their respective steering directions, simplifying the initial setup and training processes.
- ๐๐ฎ๐ญ๐จ๐ง๐จ๐ฆ๐จ๐ฎ๐ฌ ๐๐ข๐ฆ๐ฎ๐ฅ๐๐ญ๐ข๐จ๐ง: Once trained, the network can independently drive the car in the Udacity simulator, interpreting road scenarios and making steering adjustments without human input.