Your Daily AI Research tl;dr | 2022-06-05

Welcome to your official daily AI research tl;dr (and news) intended for AI professionals and enthusiasts.

In this newsletter, I share the most exciting papers I find on a daily basis, along with a short summary to help you quickly seize if the paper is worth investigating. I will also take this opportunity to share daily interesting news in the field. I hope you enjoy the format of this newsletter, and I would gladly take any feedback you have in the comments to improve it.

Now, let's get started with this iteration!

1?? Deep Learning on Implicit Neural Datasets?

As the authors shared:

"Introducing INR-Net: a framework for learning arbitrary tasks on Implicit Neural Representations!

- Agnostic to type of INR

- Universal approximator

- Continuous generalization of CNNs

INR-Net samples continuous data stored by INRs via quasi-Monte Carlo integration. It treats the INR as a black box, so it works on any type of INRs, making it more flexible than meta-learning.

In fact, it can approximate many continuous maps between integrable functions.

We design INR-Nets as a relaxation of grid-based networks to the continuous domain, enabling weight initialization with a pre-trained CNN.

INR-Nets are applicable to INR classification, segmentation, generation, and more!"

Link to the paper: https://arxiv.org/pdf/2206.01178.pdf

2?? SupMAE: Supervised Masked Autoencoders Are Efficient Vision Learners?

As we covered in a previous iteration, masking is becoming increasingly popular as a pre-training method, allowing the model to learn important features about the data to be able to reconstruct it.

This paper incorporates explicit supervision into the Masked Autoencoders (MAE) framework. SupMAE extends MAEs by adding a branch for supervised classification in parallel with the existing reconstruction objective. "In the pre-training phase, only a subset of the visible patches is processed by a ViT encoder. Their corresponding patch features are used to (1) reconstruct the missing pixels; and (2) classify the category. In the fine-tuning phase, the encoder is applied to uncorrupted images for recognition tasks."

"SupMAE is efficient and can achieve comparable performance with MAE using only 30% compute when evaluated on ImageNet with the ViT-B/16 model."

Link to the paper: https://arxiv.org/pdf/2205.14540.pdf

Code: https://github.com/cmu-enyac/supmae

?? How would I look like if my photo or video was made into a webtoon?

WebtoonMe answers this question with a very cool project and demo you can use for free.

Demo: https://webtoon.github.io/WebtoonMe/app.html

Project link: https://webtoon.github.io/WebtoonMe/en

And we are already at the end of this iteration! Please subscribe and share it with your techy friends if you've enjoyed it. Once again, let me know how to improve this format as this is something I have wanted to do for quite some time and haven't figured out the best way to do so. I hope you liked the decisions here, and I would be glad to hear from you to make it even better with time.

Thank you for reading, a fellow AI enthusiast and researcher.