From the course: TensorFlow 2.0: Working with Images
Join today to access over 23,200 courses taught by industry experts.
Why the poor performance with neural networks?
From the course: TensorFlow 2.0: Working with Images
Why the poor performance with neural networks?
- So why did our neural network perform so poorly on the CIFAR-10 dataset? Well, a neuro network doesn't take into account the spatial structure of the image. What makes a face, a face is that typically we have a pair of eyes, a nose, and mouth all close together. If we just flatten an image at the start, these details aren't captured. There are a couple of differences between the complexity of the images between the fashion-MNIST dataset and the CIFAR-10 dataset. And so let's take a look at them. With the fashion-MNIST dataset, we only have a single object in the image and all of the objects are in the center of the image. The CIFAR-10 dataset is far more realistic. There are color images with other items also part of the image and not all the images have the main object in the center. Our neural network worked well for gray scale images that have only a single channel. Now with color images we have three channels,…
Practice while you learn with exercise files
Download the files the instructor uses to teach the course. Follow along and learn by watching, listening and practicing.
Download courses and learn on the go
Watch courses on your mobile device without an internet connection. Download courses using your iOS or Android LinkedIn Learning app.