rachit-ranjan16 / Kiera

A Deep Learning Traffic Sign Recognition System

Kiera

Deep Learning Traffic Sign Recognition System

Description

A Deep Learning Application for Traffic Sign Recognition

Prerequisites

• Python 3.6+
• Ubuntu 16.04LTS
• Virtual Environment for Python with pip
  • Anaconda
  • virtualenv, pyvenv
• Network Access for getting dependencies

Setup Instructions

• Clone Repo
  • git clone https://github.com/rachit-ranjan16/Kiera.git
• Activate Virtual Environment
• Install RabbitMQ
  • sudo apt-get -y -q install rabbitmq-server
• Get project dependencies in place
  • cd Kiera
  • python setup.py develop
• Start up Django(Dev Mode)
  • python manage.py runserver 0.0.0.0:8080
• Start up Celery(Dev Mode)
  • celery -A Kiera worker -l info

Design Documentation

Deep Neural Network

• Layer 1 Input Conv
  • kernel 5x5
  • stride 1x1
  • reLU activation
• Layer 2 Conv
  • kernel 5x5
  • stride 1x1
  • reLU activation
• Layer 3 MaxPooling
  • pooling size 2x2
  • dropout 0.25
• Layer 4 Conv
  • kernel 5x5
  • stride 1x1
  • reLU activation
• Layer 5 Conv
  • kernel 5x5
  • stride 1x1
  • reLU activation
• Layer 6 MaxPooling
  • pooling size 2x2
  • dropout 0.25
• Layer 7 Flatten
• Layer 8 Fully Connected
  • dropout 0.25
  • relU activation
• Layer 9 Output
  • SoftMax Classification

Microservice Architecture

Component Diagram

Sequence Diagram