Page 193 - Ai_417_V3.0_C9_Flipbook
P. 193
Data Exploration: Explore and analyse the acquired data to understand its structure, quality, and
relationships. This phase involves data cleaning, visualisation, and statistical analysis to uncover patterns
and insights that inform subsequent modelling.
Clean and preprocess the data to extract relevant features such as plastic concentrations, ocean currents,
and marine life habitats. Explore visualisation techniques to understand spatial and temporal patterns of
plastic waste.
Modelling: Develop and implement AI models using appropriate algorithms and techniques. This
phase includes feature selection, model training, parameter tuning, and validation to optimise model
performance and accuracy.
Develop AI models such as convolutional neural networks (CNNs) for image recognition to detect and
classify plastic debris in ocean images. Implement reinforcement learning algorithms for optimising
cleanup strategies based on real-time data.
Evaluation: Assess the performance of trained AI models using evaluation metrics and validation
techniques. Validate model predictions against ground truth data to ensure reliability and generalisation
ability across different scenarios.
Evaluate model performance using different metrics to ensure accurate detection and classification of plastic
waste. Validate models with cross-validation techniques and real-world testing in different ocean regions.
Deployment: Integrate the validated AI models into the operational environment for real-world
applications. Deploy models using scalable and efficient frameworks, monitor performance
post-deployment, and iterate as necessary to maintain effectiveness and relevance.
Integrate AI models into autonomous drones or underwater robots for real-time monitoring and cleanup
operations. Implement cloud-based solutions for scalable deployment across global ocean regions.
Continuously monitor model performance and refine algorithms based on feedback from cleanup
operations and environmental assessments.
By following this AI project cycle, we can effectively leverage technology to address the critical issue of
ocean plastic pollution, leading to improved marine ecosystem.
Let us map the steps of Cleaning Ocean project to the steps in the AI project cycle.
Define Analyse data Evaluate model
objectives for to visualise accuracy and
AI-based ocean plastic waste effectiveness
plastic detection patterns and using
and removal. distributions. environmental
Data Data metrics.
Acquisition Modelling Deployment
Problem Data Data
Scoping Exploration Develop AI Evaluation
Gather satellite, models for Implement AI
drone, and accurate models in drones
oceanographic plastic waste for real-time
data on plastic detection and ocean cleanup
distribution. classification. operations.
Watch the video-The Ocean Cleanup
https://www.youtube.com/watch?v=xz21cKgRHxI
AI Reflection, Project Cycle and Ethics 191

