Ciro Santilli @cirosantilli  40

The official page: www.image-net.org/challenges/LSVRC/index.php points to a download link on Kaggle: www.kaggle.com/competitions/imagenet-object-localization-challenge/data Kaggle says that the size is 167.62 GB!

To download from Kaggle, create an API token on kaggle.com, which downloads a kaggle.json file then:The download speed is wildly server/limited and take A LOT of hours. Also, the tool does not seem able to pick up where you stopped last time.

Another download location appears to be: huggingface.co/datasets/imagenet-1k on Hugging Face, but you have to login due to their license terms. Once you login you have a very basic data explorer available: huggingface.co/datasets/imagenet-1k/viewer/default/train.

Bibliography:

Subset generators:

Unfortunately, since ImageNet is a closed standard no one can upload such pre-made subsets, forcing everybody to download the full dataset, in ImageNet1k, which is huge!

By cranks:

This is a quick presentation that goes over some of the most common difficulties people find with Git.

The fundamental insight of Git design is: a SHA represents not only current state, but also the full history due to the Merkle tree implementation, see notably:

This makes it so that you will always notice if you are overwriting history on the remote, even if you are developing from two separate local computers (or more commonly, two people in two different local computers) and therefore will never lose any work accidentally.

It is very hard to achieve that without the Merkle tree.

Consider for example the most naive approach possible of marking versions with consecutive numbers:

If Local 1 were to push to Remote first, how could Local 2 notice that when it tries to push itself? The navie method of just checking: "does Remote have commit "2"" does not work, because Local 2 has a different version of commit 2 than local 1.

This is going to be the most important application of generative AI. Especially if we ever achieve good text-to-video.

Image generators plus human ranking:

www.pornhub.com/view_video.php?viewkey=ph63c71351edece: Heavenly Bodies Part 1: Sister's Mary First Act. Pornhub title: "AI generated Hentai Story: Sexy Nun alternative World(Isekai) Stable Diffusion" Interesting concept, slide-narrated over visual novel. The question is how they managed to keep face consistency across images.

Original paper: Section "GAN paper".

2023 full connectome published on Science by Marta Zlatic's lab at the MRC:

The hard part then is how to make any predictions from it:

2024: www.nature.com/articles/d41586-024-03190-y Largest brain map ever reveals fruit fly's neurons in exquisite detail

As of 2022, it had been almost fully decoded by post mortem connectome extraction with microtome!!! 135k neurons.

That article mentions the humongous paper elifesciences.org/articles/66039 elifesciences.org/articles/66039 "A connectome of the Drosophila central complex reveals network motifs suitable for flexible navigation and context-dependent action selection" by a group from Janelia Research Campus. THe paper is so large that it makes eLife hang.