Stretch Goals: ResNet, GoogLeNet, Neural Style, GANs, emotion recognition, and more.
I'm back here today with the stretch goals for the Deep Learning for Computer Vision with Python Kickstarter campaign.
So, what are stretch goals?
I addressed the question in my previous update, but I have included the gist below for ease of reading:
"When a campaign has hit its initial funding target, the creator often creates funding goals that are higher than the original target. The creator promises that if the higher funding goals are met then they will deliver 'bonuses' to the original product."
The stretch goals for this book were determined based on the numerous emails and comments I have received since the Kickstarter campaign launched.
Take a look at the stretch goals below:
- $115,000 — More Networks: GoogLeNet and ResNet. I’ll cover more network architectures, including GoogLenet and ResNet. We’ll learn the fundamentals of “micro-architectures/modules” in the Practitioner Bundle and then apply them to CIFAR-10 and CIFAR-100. In the ImageNet Bundle I’ll demonstrate how to train both GoogLeNet and ResNet on the ImageNet dataset.
- $125,000 — Smile detection. This bonus chapter will be part of the Starter Bundle (and all higher tier bundles). I’ll treat this chapter as a “case study”, demonstrating how to train a CNN from scratch to detect and recognize smiles in images and video streams.
- $135,000 — Learning facial expressions and emotions with CNNs. Building on the previous bonus chapter, we’ll extend simple smile detection to the much more complex task of emotion recognition in images and video streams using deep learning. This chapter will be part of the ImageNet Bundle.
- $145,000 — Deep dreaming and neural style. Discover how to use deep learning to transform the artistic styles from one image to another. This bonus chapter will be part of the Practitioner Bundle and ImageNet Bundle.
- $165,000 — Generative Adversarial Neural Networks (GANs). Discover how to utilize two neural networks (a generative model and a discriminative model) to produce photorealistic images that look authentic to humans. We will cover GANs in the Practitioner Bundle and ImageNet bundle.
- $175,000 — Image super resolution with deep learning. I’ll demonstrate how to construct high-resolution images from a single, low resolution input using deep learning algorithms. The Super Resolution chapter will be part of the Practitioner Bundle and ImageNet Bundle.
If we exceed $175K in funding I will release more stretch goals as well.
If you are interested in these Kickstarter stretch goals and want to help us reach the higher target, you can simply head to the campaign page and click the Manage button next to your pledge amount.
Kickstarter also provides a FAQ on how to manage your pledge here:
From there, choose a higher level tier or simply increase your pledge amount.
Finally, if you can spare a second it would be super helpful if you could share this Kickstarter campaign link with your friends, colleagues and co-workers on your favorite social media outlet (Facebook, Twitter, LinkedIn, G+, etc.):
The more people who back the project, the more content I can include in the book!
If you have any questions regarding the stretch goals please be sure to let me know by commenting or sending me a message.
I'll be back tomorrow to discuss our first target stretch goal, "More Networks: GoogLeNet and ResNet". This discussion will also include some neat visualizations of the Inception and Residual modules.
See you then!