Progress Report-1
-
Group No: 7
-
Group Members Name and Enrollment Number:
- Shivangi Suyash (9921103053)
- Bhavya Srivastava (9921103089)
- Palak Agarwal (9921103093)
-
Supervisor: Prof. Akanksha Mehndiratta
-
Title: Making Traffic Management Smart and Secure using Federated Learning
-
Objectives of the Project:
- Develop a federated learning architecture for distributed traffic data processing and model training while ensuring data privacy.
- Implement a local data processing pipeline for preprocessing and feature extraction from traffic sensor data.
- Design and train a local machine learning model to predict optimal traffic light durations based on real-time traffic conditions such as traffic congestion and emergency vehicle presence.
- Create a secure communication protocol for model update sharing between edge devices and the central server.
- Develop an aggregation algorithm for efficiently combining model updates from multiple sources by researching and modifying available federated learning algorithms.
- Implement a system for dynamic traffic light control and emergency vehicle routing based on the federated model predictions.
-
Work Done So Far:
- Conducted a comprehensive literature review of existing traffic management systems, identifying key challenges and limitations in current approaches.
- Researched and evaluated various federated learning architectures to design the most suitable architecture for our traffic management model.
- Identified and acquired appropriate datasets for training and testing our model, ensuring diverse traffic scenarios and conditions are represented.
- Initiated preprocessing of the selected dataset, including:
- Frame Extraction: Extracted frames from the uploaded video and saved each frame as an image file.
- Compression: Compressed the saved frame images into a ZIP file for efficient storage and access.
- Object Detection: Loaded the YOLOv8 model and used it to detect and count objects in each frame for further analysis.