Edge Learning vs Deep Learning
Edge learning and deep learning are both AI-based technologies that can be used to automate a variety of applications across manufacturing and logistics operations. However, there are important differences between the two technologies, with each having its own unique characteristics.
Deep Learning – Designed for complex applications
Deep learning simulates the way neurons in the human brain strengthen and weaken connections to create an understanding of images. In deploying deep learning, neural networks are built from large image sets of similar objects. By modifying connections within and between these layers every time it is exposed to a new image, deep learning learns to identify anomalies and detect defects.
Capable of processing large and detailed image sets, deep learning is ideal for complex or highly customized applications. Such applications require advanced computational power and robust training capabilities as they introduce significant variation. To account for this variation and capture all potential outcomes, hundreds or thousands of images must be used for training. Deep learning analyzes robust image sets quickly and efficiently, delivering an effective solution for automating sophisticated tasks.
Edge Learning – Designed for ease of use
The power of AI can be applied to problems in factory automation by embedding knowledge of application requirements into the neural network connections from the start. This pre-training removes a lot of the computational load, particularly when supported by traditional machine vision tools. The end result is edge learning.
Edge learning can be trained in minutes, using as few as five to ten images. Compare this to deep learning-based solutions, which can require hours to days of training, using hundreds to thousands of images. By streamlining deployment, edge learning enables manufacturers to ramp quickly, while remaining nimble and able to adjust easily to changes.
Edge Learning vs. Deep Learning: A Quick Summary
Hear from our subject matter expert, Reto Wyss, Vice President of AI Technology at Cognex, on how edge learning meets the demand for powerful, yet simple to deploy, automation.
Our customers have always been very excited to finally solve previously unsolvable tasks. However, we also listened to their feedback about the large effort to deploy a new application. The biggest pain points are the need for many labeled images to train the model as well as the large amount of compute required. Essentially, they need a big computer with a GPU.
With the D900, we have addressed this by no longer requiring a GPU during production, but still we need it for training. With edge learning, we have now been able to push this even further. We drastically reduced the number of required training images and we dropped the need for a GPU altogether. As a matter of fact, training can be done in about one second on this tiny device itself.