Road Sign Detection and Classification Reference Implementation

Overview

The Road Sign Detection and Classification Reference Implementation monitors the exterior of a vehicle and send alerts to Cloud Dashboard. It uses a camera placed in the front of the vehicle facing forward to detect and recognize traffic signs while driving. The OpenVINO™ Semantic Segmentation model is able to detect and classify the traffic signs, and the Signs Recognition model is also implemented for this Use Case, allowing the vehicle to recognize different traffic signs placed on the road, such as speed limits, no entry, turn left or right ahead, road work, pedestrians, children crossing, no passing of heavy vehicles, etc. 

Disclaimer
This proof of concept demonstration shows how to use Edge Insights for Fleet middleware. This model is provided for demonstration purposes only. We strongly recommend developers use an appropriate model validated for their context of use. In this example, one model detects Road Signs visible to a vehicles’ front-facing camera, and another classifies/recognizes those Signs. The proof of concept demonstration also uploads JPEG snapshots of video frames containing signs to Amazon Web Services* (AWS*) cloud servers, to validate that detection events have been triggered accurately.

Select Configure & Download to download the reference implementation and the software listed below.

Configure & Download

Legal Disclaimers
Recipient is solely responsible for compliance with all applicable regulatory standards and safety, privacy, and security related requirements concerning Recipient's use of the Intel hardware and software.
Recipient is solely responsible for any and all integration tasks, functions, and performance in connection with use of the Intel hardware or software as part of a larger system. Intel does not have sufficient knowledge of any adjoining, connecting, or component parts used with or possibly impacted by the Intel hardware or software or information about operating conditions or operating environments in which the Intel hardware or software may be used by Recipient. Intel bears no responsibility, liability, or fault for any integration issues associated with the inclusion of the Intel hardware or software into a system. It is Recipient’s responsibility to design, manage, and assure safeguards to anticipate, monitor, and control component, system, quality, and or safety failures.

• Time to Complete: Approximately 60 minutes
• Programming Language: Python*
• Available Software: Intel® Distribution of OpenVINO™ toolkit 2021.4.2 Release

Recommended Hardware

The below hardware is recommended for use with this reference implementation. For other suggestions, see Recommended Hardware.

• ADLINK MXE-5500 Series

• NEXCOM VTC 7252-7C4IP

Target System Requirements

• Ubuntu* 20.04 LTS

• 6th to 10th Generation Intel® Core™ processors with Intel® Iris® Plus graphics or Intel® HD Graphics

How It Works

The reference implementation contains a full pipeline of analytics on video streams from IP cameras mounted inside a car area with a front angle and a computer using an Intel® Core™ processor or an Intel Atom® processor onboard the bus. Pretrained models are used to inference, detect and classify road signs.

This reference implementation contains a notification subsystem which includes a cloud dashboard and a cloud storage for the bus operator and fleet manager.

Figure 1: Architecture Diagram

Get Started

Step 1: Install the Reference Implementation

Select Configure & Download to download the reference implementation and then follow the steps below to install it.

NOTE: The images provided in the reference implementation are ONLY to be used for validating the accuracy of detection events.

Configure & Download

NOTE: If the host system already has Docker* images and containers, you might encounter errors while building the reference implementation packages. If you do encounter errors, refer to the Troubleshooting section at the end of this document before starting the reference implementation installation.

1. Open a new terminal, go to the downloaded folder and unzip the downloaded RI package.

   unzip road_sign_detection_and_classification.zip
   

2. Go to the road_sign_detection_and_classification/ directory.

   cd road_sign_detection_and_classification/
   

3. Change permission of the executable edgesoftware file.

   chmod 755 edgesoftware
   

4. Run the command below to install the Reference Implementation.

   ./edgesoftware install
   

5. During the installation, you will be prompted for the Product Key. The Product Key is contained in the email you received from Intel confirming your download.

   

   Figure 2: Product Key

6. When the installation is complete, you see the message "Installation of package complete" and the installation status for each module.

   

   Figure 3: Installation Success

   NOTE: If you encounter any issues, please refer to the Troubleshooting section at the end of this document. Installation failure logs will be available at the path: /var/log/esb-cli/Road_Sign_Detection_and_Classification_<version>/output.log

7. To start the application, change the directory using the cd command printed at the end of the installation process:

   cd /opt/intel/eif/EII-UseCaseManager
   

Step 2: Run the Application

Prerequisites

• Set Up ThingsBoard* Cloud Data
• Set Up Amazon Web Services* Cloud Storage

1. Run the application. Copy and run the make webui command from the end of the installation:

   make webui
   

2. Open the Web UI: Go to 127.0.0.1:9090 on your web browser.

   

   Figure 4: Reference Implementation Dashboard

3. If you installed your ThingsBoard Cloud Server and you have enabled S3 Bucket Server on your AWS account, you can provide your configured AWS Access Key ID, AWS Secret Access Key, AWS S3 Bucket Name, Thingsboard IP, Thingsboard Port and Thingsboard Device token on the Cloud Data Configuration tab. After you completed the Cloud configuration, make sure you click on Save Credentials and Save Token buttons. Now you can import the ThingsBoard dashboard as described at the end of the Set Up ThingsBoard* Cloud Data to enable all dashboard features, including the cloud storage.

   

   Figure 5: Configuration Tab Contents

   NOTE: If you don't have an AWS account, you will not be able to access Storage Cloud. You can still enable the ThingsBoard Cloud Data if you configured it locally or on another machine.

4. Access the Road Sign Detection and Classification Dashboard with the following steps.

   • Go to sidebar and select the Run Application menu option.

   • Select the RSDC tab on the General Fleet Solution dashboard.

     

     Figure 6: Select Run Application Menu Option

   • Configure the use case by selecting the video sample and the CPU or GPU device for the inference model to run on it.

   • Optionally, you can also set the simulation data that you want to use. You can choose between using the KnowGo Simulator or simply use the CSV pre-recorded simulation data.

     Model Description

     Semantic Segmentation: The OpenVINO™ Semantic Segmentation model detects and classifies the traffic signs.

     Sign Recognition: The OpenVINO™ Signs Recognition model recognizes different traffic signs placed on the road, such as speed limits, no entry, turn left or right ahead, road work, pedestrians, children crossing, no passing of heavy vehicles, etc.

     

     Figure 7: Configure Use Case

   • Click on the Browse button and search for the sample video delivered with the application at the following path: /opt/intel/eif/EII-UseCaseManager/modules/EII-RoadSignDetection-UseCase/config/VideoIngestion/test_videos/ and select the one available.

     NOTE: These images are ONLY to be used for validating the accuracy of detection events.

     

     Figure 8: Select Sample Video

   • After selecting the video sample, select the device for the inference model. Options include CPU or GPU. Click on Run Application.

   • The application will start the Visualizer App that will detect and classify the traffic signs as in the following image:

     NOTE: These images are ONLY to be used for validating the accuracy of detection events.

     

     Figure 9: Visualizer Output

5. After the visualizer starts, you can go to the ThingsBoard link and check the alerts sent by the reference implementation. If you configured the AWS credentials, you will also have access to pictures taken by the application on the video stream.

   

   Figure 10: Intel Fleet Manager Dashboard shown in ThingsBoard

6. You can also check the cloud storage from the Storage menu option.

   NOTE: These images are ONLY to be used for validating the accuracy of detection events.

   

   Figure 11: Reference Implementation Storage Menu Option

Run in Parallel with Driver Behavior Analytics Reference Implementation

To run this task you will need to download and install Driver Behavior Analytics Reference Implementation.

For more details about parallel execution, see the Edge Insights for Fleet Use Case Manager documentation.

Prerequisites

• Follow the steps to install Driver Behavior Analytics after installing Road Sign Detection and Classification

Steps to Run the Application

1. Change directory to EII-UseCaseManager on your terminal:

     cd /opt/intel/eif/EII-UseCaseManager
   

2. Run the following command to start the web server application.

     make webui
   

3. Open your browser and go to 127.0.0.1:9090.

4. Configure both installed reference implementations by setting the video source and the target (CPU, GPU or HETERO). Click on Run Application.

   NOTE: Configure each reference implementation by selecting the desired tab. For example, click the Run Application menu option, then click on RSDC to configure the Road Sign Detection and Classification RI. Next, click on DBA to configure the Driver Behavior Analytics RI.

   

   Figure 12: Configure Road Sign Detection and Classification Reference Implementation

   

   Figure 13: Configure Driver Behavior Analytics Reference Implementation

5. Wait for both Visualizers to get up and running.

   

   Figure 14: Visualizer Output for 2 Reference Implementations

NOTE: If you reinstall the first RI, you must reinstall the second one.

Summary and Next Steps

This application successfully implements Intel® Distribution of OpenVINO™ toolkit plugins to detect and classify road signs.

As a next step, try the following:

Extend the RI further to provide support for feed from network stream (RTSP camera), and the algorithm can be optimized for better performance.

Learn More

To continue your learning, see the following guides and software resources:

• Visit Edge Insights for Fleet for additional reference implementations.
• Intel® Distribution of OpenVINO™ toolkit documentation

Known Issues

Uninstall Reference Implementation

If you uninstall one of the reference implementations, you need to reinstall the other reference implementations because the Docker images will be cleared. 

Troubleshooting

Installation Failure

If the host system already has Docker images and its containers running, you will have issues during the RI installation. You must stop/force stop existing containers and images.

• To remove all stopped containers, dangling images, and unused networks:

  sudo docker system prune --volumes
  

• To stop Docker containers:

  sudo docker stop $(sudo docker ps -aq)
  

• To remove Docker containers:

  sudo docker rm $(sudo docker ps -aq)
  

• To remove all Docker images:

  sudo docker rmi -f $(sudo docker images -aq)
  

Docker Image Build Failure

If Docker image build on corporate network fails, follow the steps below.

1. Get DNS server using the command:

   nmcli dev show | grep 'IP4.DNS'
   

2. Configure Docker to use the server. Paste the line below in the /etc/docker/daemon.json file:

   { "dns": ["<dns-server-from-above-command>"]}
   

3. Restart Docker:

   sudo systemctl daemon-reload && sudo systemctl restart docker
   

Installation Failure Due to Ubuntu Timezone Setting

While building the reference implementation, if you see /etc/timezone && apt-get install -y tzdata && ln -sf /usr/share/zoneinfo/${HOST_TIME_ZONE} /etc/localtime && dpkg-reconfigure -f noninteractive tzdata' returned a non-zero code: 1 make: *** [config] Error 1

Run the following command in your terminal:

sudo timedatectl set-local-rtc 0

Installation Encoding Issue

While building the reference implementation, if you see ERROR: 'latin-1' codec can't encode character '\\u2615' in position 3: ordinal not in range(256)

Run the following command in your terminal:

export LANG=en_US.UTF-8

Can't Connect to Docker Daemon

If you can't connect to docker daemon at http+docker://localhost, run the following command in your terminal:

sudo usermod -aG docker $USER

Log out and log back in to Ubuntu.

Check before retrying to install if group Docker is available for you by running the following command in a terminal:

groups

The output should contain "docker".

Installation Timeout When Using pip or apt Commands

You may experience a timeout issue when using the People's Republic of China (PRC) internet network.

Make sure that you have a stable internet connection while installing the packages. If you experience timeouts due to Linux* apt or Python* pip installation, try to reinstall the package.

Support Forum

If you're unable to resolve your issues, contact the Support Forum.