Eirikur Agustsson

Research Scientist, Gemini Team, Google Deepmind

[Eirikur Agustsson @etagust]

Check out our latest Gemini models!

[Google DeepMind @GoogleDeepMind]

Gemini 2.0 is now available to everyone. ✨

⚡ Start using an updated 2.0 Flash in @Google AI Studio, @GoogleCloud’s #VertexAI and in @GeminiApp.

We’re also introducing:
🔵 2.0 Pro Experimental, which excels at coding.
🔵 2.0 Flash-Lite, our most cost-efficient model yet.
🔵 2.0 Flash Thinking Experimental, in @GeminiApp.
🧵 https://t.co/EvAHQ927Va

[Jeff Dean @JeffDean]

What a way to celebrate one year of incredible Gemini progress -- #1🥇across the board on overall ranking, as well as on hard prompts, coding, math, instruction following, and more, including with style control on.

Thanks to the hard work of everyone in the Gemini team and elsewhere at Google! 🎊

[Emilien Dupont @emidup]

We build neural codecs from a *single* image or video, achieving compression performance close to SOTA models trained on large datasets, while requiring ~100x fewer FLOPs for decoding ⚡ #CVPR2024

https://t.co/Y7rvJ7Mj7e

[Eirikur Agustsson @etagust]

Just the usual Monday evening in Iceland, aside from a volcanic eruption seen from the backyard...

[Eirikur Agustsson @etagust]

@BerivanISIK @GoogleAI @Google Congrats Berivan!

[Eirikur Agustsson @etagust]

Check out our latest work, showing that Finite Scalar Quantization greatly simplifies VQ based representation learning!

[Fabian Mentzer @mentzer_f]

🎉🎉🎉 Say goodbye to the VQ headache 🤕 - in our latest paper, we show a simple drop-in replacement for VQ based on scalar quantization that does not need any of the auxiliary losses and tricks from the VQ literature!

[AK @_akhaliq]

Finite Scalar Quantization: VQ-VAE Made Simple

paper page: https://t.co/BqUXMdUK1C

propose to replace vector quantization (VQ) in the latent representation of VQ-VAEs with a simple scheme termed finite scalar quantization (FSQ), where we project the VAE representation down to a few dimensions (typically less than 10). Each dimension is quantized to a small set of fixed values, leading to an (implicit) codebook given by the product of these sets. By appropriately choosing the number of dimensions and values each dimension can take, we obtain the same codebook size as in VQ. On top of such discrete representations, we can train the same models that have been trained on VQ-VAE representations. For example, autoregressive and masked transformer models for image generation, multimodal generation, and dense prediction computer vision tasks. Concretely, we employ FSQ with MaskGIT for image generation, and with UViM for depth estimation, colorization, and panoptic segmentation. Despite the much simpler design of FSQ, we obtain competitive performance in all these tasks. We emphasize that FSQ does not suffer from codebook collapse and does not need the complex machinery employed in VQ (commitment losses, codebook reseeding, code splitting, entropy penalties, etc.) to learn expressive discrete representations.

[Fabian Mentzer @mentzer_f]

1/ Excited to share that our work on masked transformers for compression has been accepted at #ICCV2023! We show two things: a) how a simple transformer based architecture can lead to state-of-the-art image compression results: https://t.co/WnB74jsqKk

[Eirikur Agustsson @etagust]

Have a look at our work on using diffusion models for SoTA generative image compression!

[Lucas Theis @lucastheis]

New paper on high-resolution image compression with diffusion models, achieving SoTA results in terms of FID:
https://t.co/9CJ6TdnVr7

[Eirikur Agustsson @etagust]

@giffmana @ArYoMo @wightmanr @rasbt @ClementDelangue @mentzer_f Assuming huffman tables are fixed, I don't se a big diff between huffman coding and translation. I think one issue is "how catastrophic" errors are. E.g. if you get one bit wrong in arithmetic coding, your decoder will just produce garbage.

[Eirikur Agustsson @etagust]

We train a decoder bridging the two regimes.
From a single representation, we can decode a low MSE reconstruction close to the input input, a realistic reconstruction with high perceptual quality, or anything in between.

Joint work with @minnend, @george_toderici and @mentzer_f!

[Eirikur Agustsson @etagust]

Excited to share our work on Multi-Realism Image Compression, which we'll present at upcoming CVPR!

https://t.co/k8iC6I9hJO

Generative compression produces synthetic details which looks realistic. But what if we want to stay close to the original?

Let's do both! (1/2)

[Fabian Mentzer @mentzer_f]

I'll present this work on Thursday in New Orleans at #NeurIPS2022 !! See u there ☀️👾

[Eirikur Agustsson @etagust]

@hashseed Unfortunately no official Google office at the moment, but I do have some colleagues that I can meet for lunch though!

[Eirikur Agustsson @etagust]

A brief life update: 🇨🇭➡️🇮🇸

After 10 wonderful years in Switzerland, I have finally returned back to my home country, Iceland.

Goodbye Google Zurich, and hello Google Iceland 🥶🌋

[Eirikur Agustsson @etagust]

@hardmaru I believe that's the right order of magnitude 😀

[Fabian Mentzer @mentzer_f]

Will be presented at #NeurIPS! 🎉🎉🎉

[Eirikur Agustsson @etagust]

This Icelandic company just reveived a 100M+€ EU grant to build a massive CO2 storage hub in my home town, Hafnarfjordur, Iceland.

Super cool tech that basically turns CO2 into stone by pumping carbonated water into lava!

[Carbfix @CarbFix]

Carbfix has been selected for a large EU grant award from the European Innovation Fund to build the Coda Terminal, the world's first large-scale CO2 transport and mineral storage hub at Straumsvík, Iceland! 💎💎💎
https://t.co/JSdxuVG79M
#Innovastionfund #EU #Carbfix

[Eirikur Agustsson @etagust]

Looking forward to present our paper on Neural Video Compression with GANs at #ECCV2022 😊

[Fabian Mentzer @mentzer_f]

Our paper on ✨Neural Video Compression with GANs✨ was accepted at #ECCV2022 🎉🥳. We just uploaded the final version to arXiv, check it out: https://t.co/0yBWEmQ3J9

[Eirikur Agustsson @etagust]

Apparently Twitter isn't confident that a tweet by the Chairman @ Twitter about the world's wealthiest person is about "Business personalities"? 😅