Image Counts For Neural Network Training

Pete Warden shares his rule of thumb for how many images you need to train a neural network:

In the early days I would reply with the technically most correct, but also useless answer of “it depends”, but over the last couple of years I’ve realized that just having a very approximate rule of thumb is useful, so here it is for posterity:

You need 1,000 representative images for each class.

Like all models, this rule is wrong but sometimes useful. In the rest of this post I’ll cover where it came from, why it’s wrong, and what it’s still good for.

Read on to learn where the number 1000 came from and get some good hints, like flipping and rescaling images.

Related Posts

Bayesian Neural Networks

Yoel Zeldes thinks about neural networks from a different perspective: The term logP(w), which represents our prior, acts as a regularization term. Choosing a Gaussian distribution with mean 0 as the prior, you’ll get the mathematical equivalence of L2 regularization. Now that we start thinking about neural networks as probabilistic creatures, we can let the fun […]

Read More

Combining Apache Kafka With TensorFlow

Kai Waehner has an example of an application which uses Apache Kafka to stream car sensor data to TensorFlow on Google ML Engine: A great benefit of Confluent MQTT Proxy is simplicity for realizing IoT scenarios without the need for a MQTT Broker. You can forward messages directly from the MQTT devices to Kafka via the MQTT […]

Read More


December 2017
« Nov Jan »