Welcome and Learning Objectives
Quick TensorFlow
Image Classification with Transfer Learning using TensorFlow 2.x
Creating and Deploying Machine Learning Models to Mobile Devices
Welcome to Quick™ TensorFlow, a series of uniquely designed lessons that quickly help you to gain mastery of Google’s easy-to-use, powerful machine learning framework: TensorFlow.
Let me tell you a little bit about this course, Image Classification with Transfer Learning.
Modern machine learning practitioners that create image recognition models rarely start with training an entire Convolutional Neural Network (CNN) from scratch. The reasons are simple: it is rare to have a dataset large enough and a computer powerful enough to achieve good results in a meaningful time frame. What we do instead is to find a model that was pre-trained on a very large dataset (e.g. ImageNet, which contains about 1.4 million images with 1000 categories), and then we re-purpose that model for the specific task we are interested in. This provides machine learning practitioners with a well trained, highly accurate artificial neural network that becomes readily available with minimal effort!
In this course, you will learn how to develop powerful machine learning models like these for image classification and object detection, and deploy those models on mobile devices. We will cover the range of topics from convolutional neural networks to transfer learning and TensorFlow Lite.
As a bonus, you will get free, early access to our new tool, Pallet, that makes deploying your models to a mobile phone take minutes instead of days or weeks!
So, take a look at the syllabus and the hands-on labs for more details on what you’ll learn and we’ll see you inside!
Syllabus and Learning Objectives
Modern Image Recognition
Using Pre-trained Convolutional Neural Networks
Data Preprocessing
Dataset Downloads and Data Preparation using TensorFlow Datasets
Creating Base Model from Pre-trained Convolutional Neural Networks
Creating the Base Model from the MobileNetV2 Model Developed at Google
Freezing the Convolutional Base
Freezing the Base Model's Weights and Biases
Feature Extraction
Creating and Testing the Feature Extraction Capabilities
Adding New Classification Layers
Wrapping the Convolutional Base in a Keras Sequential Model; Adding a New Classification Head.
Stacking Layers, Compiling, Training and Testing the Model
Loss and Optimization
Training Statistics and Predictions
Training and Validation Metrics and Predictions
Readying TensorFlow Models for Deployment to Mobile
Creating and Saving TensorFlow Lite Machine Learning Models
Deploying Models to your Android Phone
Pallet, our No-Code Model Deployment Tool! Quick and Easy Deployment of Machine Learning Models to Your Mobile Phone
Course Structure and Format:
Python experience is the only prerequisite, but familiarity w/machine learning & image classification concepts would be helpful. The course includes multiple supplementary tutorials and Colab notebooks with introductory material. (PalletML is no code!) |
This mini-course uses Google Colaboratory (Colab). Colab Notebooks are described by Google as an executable document that lets you write, run, and share code within Google Drive. If you are familiar with the Jupyter project, Colab is like a Jupyter Notebook stored in Google Drive. You also receive a free bundled course with 10+ additional tutorials and Colab notebooks that cover many topics related to image classification and transfer leaning with an excellent overview on using Colab!
Quick TensorFlow Mini-Course: Image Classification with Transfer Learning using TensorFlow |
||
| Duration: 5-7 hours | Early Bird Price: $19.99 | Enroll Now! |
| Free, 3-Months of Access to our Unique Deployment Platform, Pallet! ($30 Value! Limited Time Offer! ) | ||