Offered By: IBM
Waste Classification: Hands-on Guide to Transfer Learning
In this Guided Project, you will use transfer learning and fine-tuning to categorize waste streams based on disposal options: organic or recyclable.
Continue readingGuided Project
Deep Learning
310 Enrolled4.3
At a Glance
In this Guided Project, you will use transfer learning and fine-tuning to categorize waste streams based on disposal options: organic or recyclable.
About
You are a data science intern at a waste management service. Your manager has asked you to create a waste classification pipeline that categorizes waste streams based on disposal options: organic or recyclable. What type of model should you use?
Most popular models are difficult to train from scratch as they require huge datasets (like ImageNet), a large number of training iterations, and very heavy computing machinery. The basic features (edges, shapes) learned by early layers in a network are generalizable. While the later layers in an already trained network tend to capture features that are more particular to a specific image classification task.
You are a data science intern at a waste management service. Your manager has asked you to create a waste classification pipeline that categorizes waste streams based on disposal options: organic or recyclable. What type of model should you use?
Most popular models are difficult to train from scratch as they require huge datasets (like ImageNet), a large number of training iterations, and very heavy computing machinery. The basic features (edges, shapes) learned by early layers in a network are generalizable. While the later layers in an already trained network tend to capture features that are more particular to a specific image classification task.
Transfer learning uses the idea that if we keep the early layers of a pre-trained network, and re-train the later layers on a specific dataset, we might be able to leverage some state of that network on a related task.
A typical transfer learning workflow in Keras looks something like this:
- Initialize base model, and load pre-trained weights (e.g. ImageNet)
- "Freeze" layers in the base model by setting training = False
- Define a new model that goes on top of the output of the base model's layers.
- Train the resulting model on your data set.
In this lab, you're going to train a transfer learning model to perform this image classification task.
A Look at the Project Ahead
After completing this guided project you will be able to:
- Perform pre-processing and image augmentation on ImageGeneratorClass objects in Keras.
-
Implement transfer learning in five general steps:
- obtain a pre-trained model,
- create base model,
- freeze layers,
- train new layers on dataset,
- improve model through fine-tuning.
- Build an end-to-end VGG16-based transfer learning model for binary image classification tasks.
What You'll Need
This course mainly uses Python and JupyterLabs. Although these skills are recommended prerequisites, no prior experience is required as this Guided Project is designed for complete beginners.
This course mainly uses Python and JupyterLabs. Although these skills are recommended prerequisites, no prior experience is required as this Guided Project is designed for complete beginners.
Frequently Asked Questions
> Do I need to install any software to participate in this project?
Everything you need to complete this project will be provided to you via the Skills Network Labs and it will all be available via a standard web browser.
> What web browser should I use?
The Skills Network Labs platform works best with current versions of Chrome, Edge, Firefox, Internet Explorer, or Safari.
Estimated Effort
25 Minutes
Level
Intermediate
Skills You Will Learn
Data Science, Deep Learning, Machine Learning, Python
Language
English
Course Code
GPXX0CXEEN