What is SageMaker Canvas?
Amazon SageMaker Canvas empowers you to transform data at petabyte-scale, and build, evaluate, and deploy production-ready machine learning (ML) models without coding. It streamlines the end-to-end ML lifecycle in a unified and secure enterprise environment.
With SageMaker Canvas, you can accelerate innovation and more quickly solve business problems by democratizing ML development across all skill levels and regardless of coding expertise.
Benefits of SageMaker Canvas:
Amazon SageMaker Canvas offers a compelling suite of benefits that democratize machine learning, making it accessible to a wider range of users without requiring coding expertise. Here are some key advantages:
1. No-Code Machine Learning:
- Intuitive Visual Interface: SageMaker Canvas provides a user-friendly, point-and-click interface, eliminating the need to write any code for the entire ML lifecycle.
- Simplified Workflow: It streamlines the process of data preparation, model building, training, evaluation, and deployment into an easy-to-follow visual workflow.
2. Accessibility for Business Users:
- Empowers Non-Technical Teams: Business analysts, domain experts, and other users without deep ML knowledge can build and deploy models, directly addressing their specific business problems.
- Bridging the Gap: It reduces the reliance on data scientists and engineers for initial model development and exploration.
3. Accelerated Model Building:
- Automated Model Selection (AutoML): Canvas automatically analyzes your data and tests various machine learning algorithms to identify the best-performing models for your specific prediction task.
- Rapid Experimentation: The no-code environment allows for quick iteration and experimentation with different data and model configurations.
4. Enhanced Collaboration and Governance:
- Seamless Integration with SageMaker Studio: Models built in Canvas can be easily shared with data scientists in SageMaker Studio for further customization and advanced analysis.
- Transparency and Explainability: Canvas provides insights into model performance and offers explanations of predictions, fostering trust and understanding.
- Model Registry Integration: Models can be registered in the SageMaker Model Registry for governance, version control, and MLOps best practices.
- Collaboration Features: Facilitates collaboration across teams through model sharing and integration with other AWS services like Amazon DataZone and Amazon QuickSight.
5. End-to-End ML Lifecycle Support:
- Data Preparation: Easily connect to various data sources, visualize data, and perform data transformations without coding.
- Model Building & Training: Automatically build and train models using AutoML capabilities.
- Model Evaluation: Assess model performance with intuitive visualizations and metrics.
- Deployment: Deploy models for real-time or batch predictions with just a few clicks.
- Monitoring (via integration with other services): While Canvas itself is no-code, the deployed models can be monitored using other AWS services.
6. Cost-Effective:
- Democratization of ML: Reduces the need for large teams of specialized data scientists for every ML task.
- Efficiency Gains: Speeds up the model development and deployment process, leading to faster time-to-value.
In essence, SageMaker Canvas empowers a broader audience to harness the power of machine learning to solve business problems, explore data-driven insights, and innovate more quickly, all within a user-friendly, no-code environment
Demo:
In this demo, lets use SageMaker Canvas to build,train and deploy an image classification(dogs & cats) model without any code in just few clicks
Step 1:Create a SageMaker domain
Navigate to SageMaker AI service in AWS console and click on ‘Create a SageMaker domain’
Opt for ‘Quick setup’ for demo purpose
Once domain is ready, you can view it in ‘Domains’ section
Step 2: Open Canvas
Click on ‘My Models’
Click on ‘Create new model’
Name the model as you wish and choose ‘Image analysis’ under problem type
Step 3: Create an image dataset
Data is the lifeblood that fuels machine learning models, enabling them to learn patterns, make predictions, and drive intelligent decisions.
Kaggle is a very popular platform for data scientists and machine learning practitioners. It hosts a vast collection of datasets uploaded by the community, ranging in size and complexity.
For this demo, I have downloaded the below dataset which has labelled images of cats and dogs
Dataset consists of two folders. One folder has images for training and other folder has images for testing the model
Create a training dataset
Upload the dog images folder
Upload the cat images folder
Wait for the dataset to be ready
Step 4: Build model
As you can see below, label distribution among data is even. We have 200 dog images and 200 cat images as part of training data
Click on ‘Quick build’
It takes around 15–30 minutes for the model to be ready
Once the model is ready, you can view and analyze the training results
Step 5: Make model predictions with new data for validation
SageMaker Canvas offers two options, one for single prediction and other for batch prediction
We will go for the batch prediction and create a test dataset
From the already downloaded dataset, open the ‘test’ folder and upload 70 dog images and 70 cat images which will form the test dataset
Once test dataset is ready, click on ‘Generate Predictions’
View and analyze the prediction results
Step 5: Deploy the model
If the prediction results are satisfactory, proceed with the model deployment. You can deploy your model to SageMaker AI hosting services and get an endpoint that can be used for inference. These endpoints are fully managed and support autoscaling.
To send an inference request to a model, you invoke the endpoint that hosts it. You can invoke your endpoints using Amazon SageMaker Studio, the AWS SDKs, or the AWS CLI
And there you have it! You’ve successfully built, trained, and deployed your very own image classification model to distinguish between cats and dogs — all without writing a single line of code, thanks to the intuitive power of Amazon SageMaker Canvas.
This is just the beginning of what you can achieve with visual AI. Imagine classifying different types of flowers, identifying objects in your home, or even analyzing product quality. SageMaker Canvas opens up a world of possibilities, empowering anyone to harness the power of machine learning for image understanding. So go ahead, explore your image data and unleash your inner AI visionary!
Top comments (1)
"It’s impressive to see how far the 'no-code' experience has come for image classification. For those of you deploying these SageMaker Canvas models into production, how are you handling the model monitoring and retraining loop? I’m curious if you’re finding it seamless to integrate the Canvas-generated models into your existing MLOps pipelines, or if you’re treating these more as 'standalone' prototypes that you eventually rebuild as custom code for long-term maintenance?"