Skip to content

Developed an image classification model for Scones Unlimited to identify delivery vehicles (bicycles vs. motorcycles) to enhance routing and loading bay assignments, thereby optimizing operational efficiency.

Notifications You must be signed in to change notification settings

paschalugwu/Developing-ML-Workflows-On-AWS

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

13 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Scones Unlimited Image Classification Project

Title: Comprehensive Report on Scones Unlimited Image Classification Project

Introduction:

  1. Project Overview:

    • This project aims to build an image classification model for Scones Unlimited, a logistics company specializing in scone delivery. The model will classify images of vehicles used by delivery drivers (bicycles vs. motorcycles) to optimize routing and loading bay assignments, enhancing operational efficiency.
  2. Personal Motivation:

    • This project was chosen to apply machine learning techniques to a real-world problem in logistics. The potential to improve operational efficiency and contribute to the field of computer vision aligns with my career goals and interests in data science and machine learning.

Methodology: 3. Data Collection and Preparation:

  • The project uses data from a provided URL. The data was successfully loaded, with labels for bicycles and motorcycles identified as 8 and 48, respectively.
  • Images were saved and uploaded to an S3 bucket for processing.
  1. Exploratory Data Analysis (EDA):
    • No additional EDA were made.

Modeling and Implementation: 5. Model Selection:

  • The project involved setting up an image classification model using SageMaker's "image-classification" algorithm. The model was trained using an ml.p3.2xlarge instance type.
  1. Implementation Details:
    • SageMaker Studio was set up with a kernel to run the project.
    • Data preparation was completed using the ETL (extract, transform, load) process.
    • The model was trained without errors, achieving a validation accuracy of 0.86 with 30 epochs.
    • The trained model was deployed, and a unique model endpoint was printed for making predictions.
    • Predictions were successfully made using a sample image.

Results and Evaluation: 7. Model Performance:

  • The model achieved a validation accuracy of 0.86.
  • A unique endpoint was used to make predictions with the deployed model.
  1. Business Impact:
    • The model can significantly improve operational efficiency for Scones Unlimited by optimizing routing and loading bay assignments based on vehicle type. This can lead to cost savings and better resource allocation.

Challenges and Solutions: 9. Obstacles Encountered:

  • Challenges included setting up the SageMaker environment and ensuring proper data handling.
  • These were addressed by following detailed setup instructions and validating each step of the process.

Conclusion and Future Work: 10. Project Summary: - The project successfully built and deployed an image classification model using AWS services. The model's performance demonstrates the potential to enhance Scones Unlimited's logistics operations.

  1. Future Improvements:
    • Future work could explore additional data sets and algorithms to further improve model accuracy.
    • Implementing data augmentation techniques and exploring more efficient ways to batch process large datasets could enhance performance.

Personal Reflection: 12. Skills and Growth: - This project enhanced my skills in using AWS SageMaker, Lambda functions, and Step Functions. It also provided valuable experience in deploying and monitoring machine learning models in a cloud environment.

  1. Conclusion:
    • I am enthusiastic about the potential of machine learning to solve real-world problems and am grateful for the opportunity to work on this project. I look forward to continuing my growth in the field of data science and machine learning.

Attachments and References: 14. Supporting Documents: - The project includes Python scripts for the Lambda functions, JSON export defining the Step Function, and screenshots of the working Step Function. - References include links to AWS documentation and articles provided by the project reviewer.

  1. References:
    • Serverless Orchestration of Distributed Workflow at Scale Using AWS Step Functions: Link
    • Deploying a Multi-Model Inference Service With AWS Lambda, Synchronous Express Workflows, Amazon API Gateway, and CDK: Link
    • A Model Is for Life, Not Just for Christmas!: Link

About

Developed an image classification model for Scones Unlimited to identify delivery vehicles (bicycles vs. motorcycles) to enhance routing and loading bay assignments, thereby optimizing operational efficiency.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published