This project was offeered to us by Dr. Harsh Verma of Glocol Networks. Glocol Networks is an Internet-of-Things (IoT) start-up with a mission of transforming how IOT, Smart Objects, and Wearables can change our lives in Smart Cities with a new paradigm to smart living and saving lives.
The purpose of this project is to create a flexible, non-invasive way of getting the current traffic light statuses of intersections.
This would allow further connectivity in future plans for smart cities, as well as be beneficial for self-driving cars and other autonomous vehicles.
This will be done by developing an AI/ML model using Intel OpenVINO as a Visual Approach for capturing Signal Change information at Traffic Intersections without interfering or touching the Traffic Controller. We are using Roboflow to train our model.
We also developed a website for displaying the change of traffic signal phases.
To test the developed model, navigate to the URL here and choose either the URL or file upload methods. Then ensure the Labels section has “On” selected. Ensure the inference result section has “Image” selected and click “Run Inference” once a valid photo URL or photo image has been uploaded.
To test the model locally, follow the instructions given in the Maintenance Manual to download all necessary files. Then, follow the instructions in the System Test Report to set up the test environment.
To test the model on Google Colab/Jupyter Notebook, follow the instructions in the System Test Report to set up the test environment.
This service is deployed automatically when a commit is merged to the Signal-Sense repository.
Open a terminal. Navigate to UI/src. Verify you have node already installed. Run "npm i" and then "node server.js". Navigate to localhost:3000 on a browser.
-
Run ‘python -m venv yolov7env’ to create a virtual environment
-
Navigate to yolov7env/Scripts and run ‘activate’
-
Run
git clone https://github.com/AmaanEziz/SignalSenseto install the model files -
Navigate to yolov7
-
Run
pip install -r requirements.txtto install the required packages -
Run ‘python detect.py --weights best.pt --conf 0.4 --img-size 640 –source ./inference/images’ to run the inference using the PyTorch model.
-
Run ‘python detect-ONNX.py’ to run the inference using the ONNX model
The model should now be fully set up and ready for detection.
Please refer to the User Guide for more information on running detection with preconfigured scripts, or https://github.com/WongKinYiu/yolov7 for more information on the model itself.
The trained model (.pt and .onnx) is stored in the https://drive.google.com/drive/folders/12Hpy5GGVG6ktBY3NA8FjU4FjZliUXtok?usp=share_link
-
Choose ‘Runtime’, your OS type, version ‘2022.3’, distribution ‘PIP’ and follow the installation guide given on the page
*Suggested to use Python version 3.9
-
Navigate to openvino_env
-
Download ‘detect-ir.py’ from this repository, the trained model from the google drive, and place in openvino_env directory
-
Create a text file named
class_labels.txt -
Navigate to openvino_env/images and place the image(s) you want to run inference on
-
Navigate back to openvino_env
-
Run
python detect-ir.pyto run inference
Please refer to the User Guide for more information on running detection with preconfigured scripts
The trained model (.xml and .bin) is stored in the https://drive.google.com/drive/folders/1NQ2DdxkQuzjxjA4i1zIZEQszl-OMXn-5?usp=share_link
| Name | Area of Focus |
|:--------------:|:---------------------------:|
| Manvir Kaur | Team Lead & AI/ML |
| Mindy Cha | AI/ML |
| Jian Chen | UI |
| Liana Coyle | AI/ML |
| Amaan Eziz | UI |
| Anh Duy Phan | UI |
| Luda Salova | AI/ML |
| Kyle Barreras | AI/ML |
User Manual (Work is in progress for the documentation! Links will be updated soon!): Instructions on how to run the model and how to use the website.
Maintenance Manual (Work is in progress for the documentation! Links will be updated soon!): Instructions on how to maintain the project.