DEV Community: Kushagra Garg

Enhanced Maritime Safety Through Object Detection

Kushagra Garg — Sun, 10 Nov 2024 19:43:47 +0000

Enhanced Maritime Safety Through YOLO-Based Object Detection

Introduction

Maritime safety is an ever-evolving field, with security and rescue operations being critical for those at sea. Traditional surveillance methods often face challenges such as slow response times and limited accuracy, especially in harsh maritime environments. This article introduces a YOLO-based object detection approach that leverages real-time data to identify critical objects—like humans, vessels, and maritime infrastructure—enhancing situational awareness, improving safety, and supporting quicker rescue operations.

Objectives

The main goals of this project are to:

Develop Real-Time Object Detection: Create a fast and reliable object detection system tailored for maritime settings.
Enhance Situational Awareness: Equip operators with accurate, real-time data on objects and potential risks.
Facilitate Swift Rescue Operations: Enable faster, more targeted rescue efforts for individuals in distress.
Automate Maritime Surveillance: Provide a system that autonomously monitors maritime areas, ensuring constant safety and security.

Limitations in the Existing Literature

Existing object detection methods for maritime surveillance, such as Support Vector Machines (SVM) and Convolutional Neural Networks (CNN), are well-established but often fall short in real-time adaptability. Traditional methods face challenges in low frame rates, accuracy, and adapting to various maritime conditions. This often leads to:

Delayed Detection: Low frame rates hinder real-time responsiveness.
Inaccurate Recognition: False detections or missed objects, especially in complex maritime scenarios.
Limited Adaptability: Difficulty in adjusting to changing weather, lighting, and background variations.

Why YOLO for Maritime Object Detection?

YOLO (You Only Look Once) is a popular object detection framework known for its high speed and accuracy, making it ideal for real-time applications. Its strengths in the context of maritime safety include:

High Processing Speed: YOLO processes images in a single forward pass, ideal for fast-paced maritime environments.
Enhanced Accuracy: The YOLO architecture is designed to reduce false positives and improve precision.
Generalization Across Scenarios: YOLO’s adaptability is advantageous for diverse maritime conditions, handling variations in weather, lighting, and sea states.
Computational Efficiency: YOLO’s design allows it to run efficiently on limited hardware, a benefit for remote or mobile deployments.

Dataset and Methodology

Dataset

For this project, a specialized maritime dataset was used. Key attributes include:

Image Source: Aerial imagery from drones and other platforms.
Categories: Humans, yachts, boats, jet skis, docks.
Data Annotations: Bounding boxes with class labels for each object.
Diversity: Thousands of images representing varying object sizes, orientations, and weather conditions.

Methodology

The development of the YOLO-based detection model involved several steps:

Data Preparation: Curating and annotating a robust maritime dataset.
Model Training: Using YOLOv8, the latest YOLO variant, for high-speed detection.
Model Evaluation: Testing the model on key metrics like precision, recall, and F1 score to ensure accuracy and reliability.
Deployment: Integrating the trained model into a real-time platform for maritime surveillance and rescue operations.

Detection Images

Experimental Results

The YOLOv8 model showed promising results:

High Precision and Recall: Consistent detection of various maritime objects across different conditions.
Real-Time Processing: The model achieved up to 60 FPS, enabling seamless real-time monitoring.
Improved Situational Awareness: Accurate detection in challenging scenarios, such as foggy conditions or cluttered backgrounds.
Confusion Matrix:

Metrics:

-Result:

Benefits and Impact

The deployment of this YOLO-based object detection system can bring multiple benefits to maritime operations:

Faster Rescue Response: Early detection of distressed vessels or individuals, allowing timely intervention.
Enhanced Security: Real-time surveillance and detection of unauthorized vessels or activities contribute to maritime safety.
Environmental Protection: Early detection of potential environmental hazards, such as oil spills, for proactive ecosystem protection.
Operational Efficiency: By automating object detection and monitoring tasks, the system reduces the workload on maritime operators.

Future Directions

To further improve the system, future work could focus on:

Advanced Object Recognition: Identifying specific vessel types or individuals.
Real-Time Tracking: Predicting trajectories to avoid potential collisions.
AI-Driven Decision Support: Integrating with maritime navigation and communication systems to support operators during emergencies.

Conclusion

This project highlights the potential of YOLOv8 for enhancing maritime safety through real-time object detection. By providing accurate, efficient, and adaptable object identification, the system enables faster rescue operations, boosts security, and supports proactive environmental protection. YOLO’s computational efficiency also opens doors to mobile and remote maritime monitoring applications. With continued advancements, AI-driven object detection has a promising future in safeguarding our seas.

Community Central - AWS Amplify

Kushagra Garg — Mon, 27 May 2024 01:01:53 +0000

This is a submission for the The AWS Amplify Fullstack TypeScript Challenge

What We Built

Community Central is a dynamic platform where users can create, view, update, and delete communities. Each community can have its own set of announcements, which are created by the community owners and are visible on the community page. The About tab provides detailed information about what the community is about, helping users understand its purpose and goals. Additionally, we have integrated Crisp support to ensure customer satisfaction. This project was built using Next.js for the frontend and AWS Amplify Gen 2 for the backend.

Demo and Code

Live Deployment - https://main.d7jv2qcnifbzj.amplifyapp.com
Source Code - https://github.com/Kushagra102/amplify-aws

Integrations

Data - A nested(complex) production ready CRUD.
Authentication - Via Authenticator component and email verification based Auth.
Serverless Functions - Used for sending Welcome Message via Email to the New Users.
File Storage - Used for storing Image Files throughout the application.

Journey

Our journey with Community Central started with setting up AWS Amplify Gen 2. Having experience with Gen 1, transitioning to Gen 2 was a seamless and exciting experience for us. The new features and improvements in Gen 2 significantly enhanced our development process.

The first major learning curve was establishing relationships in the datastore using the hasMany and belongsTo directives, which allowed us to query nested data efficiently. Mastering this was crucial for building a robust backend for our app.

Our initial success came with establishing authentication and creating a clutter-free CRUD interface for our users. Using Amplify UI components and connected forms was a thrilling experience. The way Amplify automatically generates create and update forms streamlined our development process, making it faster and more efficient.

We also brainstormed potential triggers for our app, such as sending a welcome email upon signup, notifying owners when a community is created, and sending announcements to community members. We successfully implemented the signup triggers and are planning to incorporate create and announcement triggers in the near future.

As for optimizing the UI, we considered using the TanStack Query approach but decided to leave this for future projects with Amplify. Our focus remains on delivering a smooth and user-friendly experience for our community platform.

Looking back, we're particularly proud of how we leveraged AWS Amplify's capabilities to build a functional and efficient application. Our next steps include implementing the remaining triggers and continuing to optimize and enhance our platform based on 100's of features that we have in our mind.

Connected Components and Feature Full

We are excited to convey that we have Incorporated both Amplify connected components and all four integrations.

1. Amplify Connected Components and Amplify UI -

Use Of Connected forms for Create and Update requests.
Use of Authenticator for Authentication.
Use of Tabs for seamless navigation.
Use of Account Setting Change Password for easy reseting of passwords.
Use of Storage Image and Manager to efficiently show and upload images to S3 (Storage).
Encorporated Connected Forms together with Tabs for creating and managing communities and announcements.
Alert was use to notify and block non-owners to change community settings.
Use of Search Field to incorporate seamless search of multiple communities.
Flex, Image, Containers, View, Divider, Heading, Button, Input, Grid and Cards.

2. Feature Full -

We have used Connected Components and all four features: Data, Authentication, Serverless functions, and File Storage.

Data: We have used DynamoDB to securely store all the information records of the communities and users. We also created complex data relationships between the tables to query and update the information efficiently.
Authentication: The users are authenticated using Amplify Auth (Cognito) to access the communities. We also save information of users like preferred usernames and names. Also, the app sends a welcome email to the user on successfully registering
File Storage: Banners are a major part of the communities to express themselves, we used S3 with amplify which made our task very easy to upload and query the image files.
Serverless: Developing for user experience is a major key point, in our app, we enhance the user experience by Sending Welcome emails using the Serverless Lambda function attached to the trigger on the Auth Service.

This is a submission by a team of 2.

Team Members:
@geoffreyanto12
@kushagra102