VehicleVision

Project Overview

The core objective of VehicleVision is to craft an image classification model that excels at differentiating between bicycles and motorcycles.

Project Objectives

1. Precise Image Classification: Engineer a robust model capable of accurately categorizing images as bicycles or motorcycles.

2. Scalable Deployment: Utilize AWS Sagemaker to deploy the model in a scalable manner, accommodating varying demand.

3. Automated Workflow: Develop AWS Lambda functions to streamline data preprocessing and orchestrate their execution using AWS Step Functions.

4. Thorough Testing: Construct a comprehensive testing and evaluation framework to ensure both the model and the workflow's dependability.

5. Monitoring and Maintenance: Implement mechanisms to actively monitor model performance and identify potential anomalies.

AWS Services Used

1. AWS Sagemaker: Leveraged for model training, deployment and model monitoring, enabling scalable and efficient machine learning operations.

2. AWS Lambda: Utilized to create serverless functions for serializing images, classify images, and result filtering.

3. AWS Step Functions: Employed to seamlessly orchestrate the execution of Lambda functions, creating a coherent and automated workflow.

4. AWS S3: Utilized as a storage solution for data staging, model artifacts, and intermediate outputs during various project phases.

Project Phases

Phase 1: Data Preparation

Prepare the dataset for model training:

1. Extract data from a designated source.
2. Transform data into a suitable format for training.
3. Load processed data into a suitable storage system.

Phase 2: Model Training and Deployment

Train and deploy the image classification model:

1. Utilize AWS's image classification algorithm for model training.
2. Deploy the trained model to AWS Sagemaker endpoint.
3. Configure AWS Model Monitor to track deployment performance.

Phase 3: Lambda Functions and Workflow Orchestration

Develop AWS Lambda functions and orchestrate their execution:

1. Create three distinct AWS Lambda functions:
   1. Serialize image (serializeImage.py)
   2. Classify image (classifyImage.py)
   3. Result filtering (filterInferences.py).
2. Design a workflow using AWS Step Functions to coordinate these functions (stepFunction.json).

• Step Function Workflow:

Phase 4: Testing and Evaluation

Thoroughly assess the workflow's effectiveness:

1. Invoke the step function with test data.
2. Validate successful and unsuccessful workflow executions.
3. Utilize SageMaker Model Monitor insights to visualize model behavior.