22.02.2024 Views

Daniel Voigt Godoy - Deep Learning with PyTorch Step-by-Step A Beginner’s Guide-leanpub

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

corresponding trained weights (the pre-trained model) are released. Finally,

everyone else can use these weights as a starting point and fine-tune them

further for a different (but similar) purpose.

That’s transfer learning in a nutshell. It all started with computer vision models

and…

ImageNet

ImageNet is an image database organized according to the WordNet [107]

hierarchy (currently only the nouns), in which each node of the hierarchy is

depicted by hundreds and thousands of images. Currently we have an

average of over five hundred images per node. We hope ImageNet will

become a useful resource for researchers, educators, students and all of you

who share our passion for pictures.

Source: ImageNet [108]

ImageNet is a comprehensive database of images spanning 27 high-level

categories, more than 20,000 sub-categories, and more than 14 million images

(check its statistics here. [109] ) The images themselves cannot be downloaded from

its website, because ImageNet does not own the copyright of these images. It does

provide the URLs for all images, though.

As you can probably guess, classifying these images was a monumental task in the

early 2010s. No wonder they created a challenge…

ImageNet Large Scale Visual Recognition

Challenge (ILSVRC)

The ImageNet Large Scale Visual Recognition Challenge (ILSVRC) evaluates

algorithms for object detection and image classification at large scale.

Source: ILSVRC [110]

The ILSVRC ran for eight years, from 2010 to 2017. Many architectures we take

for granted today were developed to tackle this challenge: AlexNet, VGG,

Inception, ResNet, and more. We’re focusing on the years of 2012, 2014, and 2015

only.

500 | Chapter 7: Transfer Learning

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!