Last Week in Computer Vision #19

Reflecting on Computer Vision and AI advancements in 2022

Author

Dasha Gurova
January 09, 2023

We've officially landed in 2023! With a fresh start comes a chance to look back and reflect on all the incredible advancements of 2022. I also highlighted some notable papers from December. Hope you will enjoy it! 🤓

In this issue of The Ground Truth: Top 10 Papers by Sebastian Raschka; Computer Vision Trends Shaping 2022; A Year of Generative Models; Book recommendations from Chip Huyen; Hinton's Forward Forward Algorithm; ViTs for MobileNet size; "Life vs. ImageNet: What I wish I had known before deploying computer vision to the real world" and more.

Author Picks 🤓

A perfect blend of predictions on MLOps, chips, and foundation models in the form of a story by Daniel Jeffries, CIO at Stability AI.

"The great acceleration of AI started with a collapse. It was 2023 and we saw a big collapse and consolidation of MLOps companies. It happened because in the last decade, Big Tech yanked AI out of the labs and universities and into the real world."

Reflections on 2022 👀

A Big Year For AISebastian Raschka shared a review of the top ten papers he has read in 2022.

And a list of Raschka's favorite 10 open-source releases that he discovered, used, or contributed to in 2022.

The Year of Disruption: The Top Computer Vision Trends Shaping 2022Analytics Vidhya has shortlisted the top 10 most useful, powerful, and trending hit list of topics in Computer vision in 2022.

The state of AI in 2022 - McKinsey Report"The results of this year’s McKinsey Global Survey on AI show the expansion of the technology’s use since we began tracking it five years ago, but with a nuanced picture underneath."

2022 Top Papers in AI — A Year of Generative ModelsThis article is a take on the 20 most impactful AI papers of 2022.

Chip Huyen: Books that made me think (as an engineer)"I wouldn’t say that I’m a great engineer today, but without these books, I would’ve been much worse." 

Computer Vision Landscape 2022 ReportThe computer vision market is expected to grow to US$17.4 billion by 2023. What exciting trends can we expect to see next year?

December Papers 🎓

Geoffrey Hinton's experiments found that the FF algorithm had a 1.4 percent test error rate on the MNIST which is just as effective as backpropagation.

OpenAI open sourced Point-E, a machine learning system that creates a 3D object given a text prompt. According to a paper published alongside the code base, Point-E can produce 3D models in one to two minutes on a single Nvidia V100 GPU.

Can Vision Transformer models run as fast as MobileNet and maintain a similar size? The paper revisits the design choices of ViTs and proposes an improved supernet with low latency and high parameter efficiency.

"In this paper, we explore the underlying differences between ViTs and CNNs, and we find that transformers detect image background features, just like their convolutional counterparts, but their predictions depend far less on high-frequency information."

Event Alert 🚨

My friends at Lakera are bringing ML developers together to share their journeys, struggles, and insights from deploying some of the most exciting computer vision applications to the real world. Tune in on Feb 15th!

From the author: My name is Dasha, I am a computer vision explorer & Community Manager at Superb AI. I am still figuring this newsletter thing out so your feedback would help a lot: Feedback Form! 👩‍💻 

Share The Ground Truth Newsletter 🤗

If you enjoy this newsletter, please forward the Subscription Link to your friends. 

Join the conversation

or to participate.