big_vision VS immich

I think this is based off Google research https://github.com/google-research/big_vision

The the vision training models are available here: https://github.com/google-research/big_vision/tree/main which I am assuming, based on the research paper is what was used for the project.

Found relevant code at https://github.com/google-research/big_vision + all code implementations here

Official Github repository: https://github.com/google-research/big_vision

An open-source implementation of the Better plain ViT baselines for ImageNet-1k research paper in Google's JAX and Flax.

An update from some of the same authors of the original paper proposes simplifications to ViT that allows it to train faster and better.

Among these simplifications include 2d sinusoidal positional embedding, global average pooling (no CLS token), no dropout, batch sizes of 1024 rather than 4096, and use of RandAugment and MixUp augmentations. They also show that a simple linear at the end is not significantly worse than the original MLP head.

Simple ViT Research Paper: https://arxiv.org/abs/2205.01580

Official Github repository: https://github.com/google-research/big_vision

Developer updates can be found on: https://twitter.com/EnricoShippole

In collaboration with Dr. Phil 'Lucid' Wang: https://github.com/lucidrains

It really is hard to leave Gmail when all of your data has been conveniently stored therein. This is one of Google's retention strategies and it is indeed brilliant.

That said, there's a vast number of self-hosted alternatives like Stalwart Mail (email) [1], Immich (images) [2], NextCloud (Google Docs) [3], etc.

[1] https://stalwa.rt

[2] https://immich.app

[3] https://nextcloud.com/

Last year we added CLIP-based image search to https://immich.app/ and even though I have a pretty good understanding of how it works, it still blows my mind damn near every day. It's the closest thing to magic I've ever seen.

I’m a big fan of https://immich.app/ and I use it every day for thousands of assets

Real-world applications often require complex queries that go beyond simple Approximate Nearest Neighbor (ANN) search. To explore a practical example of such applications, let's take a closer look at immich, a self-hosted photo and video backup solution that highlights the importance of advanced vector and traditional relational queries.

I discovered these 3 amazing projects recently:

Cryptpad, essentially google docs/sheets/forms e2e encrypted. It does include collaboration. https://github.com/cryptpad/cryptpad

Immich, google photos self hostable, with share options https://github.com/immich-app/immich

Nginxproxymanager manages certificates and proxies to self hosted stuff through nginx https://github.com/NginxProxyManager/nginx-proxy-manager

Great self hosting stuff!

I realize it's very hard, but can we maybe reconsider opening the encryption-at-rest feature request? https://github.com/immich-app/immich/issues/450

Maybe we can give temporary access to processing steps in the pipeline, then have Immich forget the keys after it does the processing?