Image recognition

As explained in Grokking Machine Learning

The computational task of identifying or classifying objects, patterns, or other content in images.

Image recognition is classifying what an image contains. For example, a model can use image pixels as features to decide whether a picture contains a dog. section 5.5

A perceptron can perform reasonably on curated image datasets such as MNIST, which contains handwritten digits, but struggles with more complicated images. Combining perceptrons into a multilayer neural network supports more powerful classification. section 5.5

A neural-network image-classification example uses MNIST:

• Flatten each 28-by-28 image into a vector and convert its digit label into a 10-entry categorical vector. section 10.3
• Train a Keras network with two hidden layers to classify the digits, then use its predictions on test images. section 10.3
• Inspect results: the network classifies some digits correctly and mistakes others, such as a 3 predicted as an 8. section 10.3

Figure 10.23 Some examples of handwritten digits in MNIST with their labels

See handwritten digit images alongside their labels. section 10.3

How Grokking Machine Learning covers it

How deep it goes

Defines pixel-based classification and demonstrates MNIST preprocessing, training, and prediction, while noting perceptron limitations.

Examples

Uses dog detection and handwritten-digit classification, including a 3 misclassified as an 8.

What you can do after

Apply via Keras by preprocessing MNIST images and labels, training a neural network, and predicting digit classes.

How it connects

Connects image recognition to classification, perceptrons, neural networks, and MNIST.

liveBookmanning.com

Read Grokking Machine Learning on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  The AI Pocket BookThe task of assigning an image to one of several possible classes based on its visual content.

Compare how 2 books explain image recognition →

Related concepts

• MNIST datasetuses · MNIST is a labeled dataset of 28-by-28-pixel grayscale images representing handwritten digits from zero through nine.
• Geminiimplemented by · Gemini is Google's family of multimodal generative AI models and the conversational AI services built around them.
• GPT-4oimplemented by · GPT-4o is an OpenAI multimodal generative model in the GPT family that processes and generates text and other supported modalities.
• perceptronused by · A perceptron is a single-layer computational model and linear classifier that combines weighted inputs with a bias and applies a threshold or activation rule.
• clusteringused by · Clustering is the technique of grouping data items according to similarity or shared characteristics.
• deep learninguses · Deep learning is a machine-learning approach that uses neural networks with multiple computational layers to learn representations from data.

This explainer was generated by Manning's knowledge graph from the text of Grokking Machine Learning. The section links point to the parts of the book it draws on.