Elastic Video Analysis Application

Description

This project focuses on developing an elastic application for video analysis using AWS Lambda and other AWS services. The application is designed to automatically scale out and in on-demand, ensuring cost-effectiveness and efficient resource utilization. It offers a multi-stage pipeline for processing user-uploaded videos, incorporating video splitting and face recognition functionalities.

How It Works

1. User Uploads Video: Users can upload videos to the application through a user-friendly interface.
2. Video Splitting (Stage 1): Upon video upload, the application splits the video into individual frames using FFmpeg. Each frame is processed separately to facilitate efficient face recognition.
3. Face Recognition (Stage 2): The extracted frames undergo face recognition using a pre-trained Convolutional Neural Network (CNN) model, specifically ResNet-34. This model is adept at detecting and recognizing faces with high accuracy.
4. Output Storage: The results of face recognition, including the names of recognized faces, are stored in an output bucket for easy access and retrieval.

Components

• Input Bucket: Stores user-uploaded videos awaiting processing.
• Stage-1 Bucket: Intermediate storage for frames generated during video splitting.
• Output Bucket: Stores the results of face recognition, including the names of recognized faces.
• Video-splitting Function: AWS Lambda function responsible for splitting uploaded videos into frames and triggering face recognition.
• Face-recognition Function: AWS Lambda function that performs face detection and recognition on individual frames.

Project Diagram

Technologies Used

• AWS Lambda: Serverless computing service for executing code in response to triggers without provisioning or managing servers.
• AWS S3: Object storage service used for storing and retrieving data.
• FFmpeg: Open-source multimedia framework for handling video, audio, and other multimedia files.
• ResNet-34: Pre-trained Convolutional Neural Network model known for its effectiveness in image classification tasks, including face recognition.

Setup

1. AWS Configuration: Set up an AWS account and configure Lambda functions, S3 buckets, and necessary permissions.
2. Lambda Function Implementation: Implement the video splitting and face recognition logic in the Lambda functions. Ensure proper error handling and logging for debugging and monitoring purposes.
3. Trigger Configuration: Configure triggers to initiate the processing pipeline upon video upload to the input bucket.
4. Testing and Monitoring: Test the application thoroughly using sample videos and monitor its performance and resource utilization using AWS CloudWatch or other monitoring tools.

Usage

1. Upload Video: Users can upload videos through the application interface.
2. Monitor Progress: Track the progress of video processing through Lambda function logs and monitoring dashboards.
3. Retrieve Results: Once processing is complete, users can retrieve the results from the output bucket, which may include text files with the names of recognized faces.

Acknowledgments

• This project was developed as part of the Cloud Computing course and serves as a practical application of serverless computing concepts.
• Special thanks to the AWS Lambda and S3 teams for providing robust and scalable services that facilitate the development of elastic applications.

License

This project is licensed under the terms of the MIT license. You are free to use, modify, and distribute the code as per the MIT license agreement.