Page 97 - Ai_C10_Flipbook
P. 97

• Goal: Address the problem of incorrect waste disposal, which leads to ineffective recycling efforts at the school.
                   The  objective  is  to  create  a  system  that  can  identify  different  types  of  waste  and  sort  them  correctly  into
                   recycling, compost, or trash bins.
                    • 4Ws Problem Canvas: To understand the step-by-step process of problem scoping, we use a method called the
                   4Ws Problem Canvas. This approach helps us identify four critical parameters that are essential for solving a
                   problem effectively.

                                         Who?             What?            Where?            Why?

                  Stage 2   Data Acquisition

                 The foundation of any AI model lies in acquiring relevant and accurate data.
                    • Data Collection: Gather images of different types of waste items, such as paper, plastic bottles, food scraps,
                   and general trash.
                    • Sources: Collect data using cameras installed near waste bins or manually label images collected by students
                   for training.
                    • Data Types: Images (photos of waste items) with labelled categories (recyclable, compostable, trash).

                  Stage 3   Data Exploration

                 Understanding the data is crucial to building an effective model.
                    • Analyse the Data: Visualise the collected images and check if there are clear differences between categories
                   like plastic, paper, food waste, and non-recyclables.
                    • Insights: Use data visualisation to understand which waste type is often confused, such as compostable versus
                   non-compostable items, and identify the features that distinguish them.
                  Stage 4   Modelling

                 Designing the model requires choosing the right algorithms and training approaches.
                    • Algorithm selection:  Choose  a  machine  learning  algorithm  suitable  for  image  classification,  such  as  a
                   Convolutional Neural Network (CNN), which is effective in processing and recognising images.
                    • Model building: Develop different models using labelled images to train the AI to recognise and classify waste
                   correctly.

                  Stage 5   Evaluation
                 Thorough evaluation ensures the reliability and accuracy of the AI model.
                    • Test models: Test multiple models to find which one distinguishes between various types of waste with the
                   highest accuracy.
                    • Validation: Test the model with new images to see if model can correctly identify items in real-time and sort
                   them into the appropriate bins.
                    • Improvement:  If  the  model  struggles  with  certain  items,  add  more  training  data  or  adjust  the  algorithm
                   parameters to improve accuracy.

                  Stage 6   Deployment
                 Deploying the model brings the AI system into real- world application.
                    • Model Integration: Integrate the model into a system where a camera and display screen are installed above
                   waste bins. The system automatically suggests which bin to use for each item based on camera detection.
                    • Monitoring: Continuously track the system's performance, making sure it correctly identifies items and reduces
                   sorting errors.
                    • Maintenance: Regularly update the model with new data to adapt to changes in waste types or introduce new
                   categories as needed.

                                                            Revisiting AI Project Cycle & Ethical Frameworks for AI   95
   92   93   94   95   96   97   98   99   100   101   102