TensorFlow 2.0: Working with Images

40mIntermediate2022-02-01

Authors

Jonathan Fernandes

Consultant focusing on data science, AI, and big data

Course details

TensorFlow 2.0 is quickly becoming one of the most popular deep learning frameworks and a must-have skill in your artificial intelligence toolkit. Using a hands-on approach, machine learning and AI model expert Jonathan Fernandes shows you the basics of working with images—both grayscale and color—in TensorFlow, and explores transfer learning and other training enhancements such as ModelCheckpoint, EarlyStopping, and TensorBoard.

Learning objectives
Differentiate between EarlyStopping and ModelCheckpoint callbacks.
Recognize how a model’s input layer and the associated dataset are related.
Explain why neural networks do not take spatial structure into account.
Describe what transfer learning is and why it is so useful.
Recognize the difference between pretrained models and fine-tuning them.
Explain the use case for EarlyStopping.

Skills covered

TensorFlowNeural Networks and Deep LearningPythonGoogleArtificial Intelligence (AI)Open SourceDeep Dive (X:Y)

Concepts

0. Introduction

01 - Work with gray and color images using transfer learning and fine-tuning
02 - What you should know
03 - What is TensorFlow

1. Neural Networks and Images

04 - Review of neural networks
05 - Working with color images and neural networks
06 - Challenge - Experiment with hyperparameters
07 - Solution - Experiment with hyperparameters

2. Transfer Learning

08 - Why the poor performance with neural networks
09 - TensorFlow Hub
10 - What is transfer learning
11 - Transfer learning with TensorFlow Hub
12 - TensorFlow Hub for CIFAR-10

3. Monitoring the Training Process

13 - Monitoring the training process
14 - Using ModelCheckpoint
15 - Working with EarlyStopping
16 - Using TensorBoard

Conclusion

17 - Next steps