Make-A-Video with Stable Diffusion 2d backbone is coming along.
Demo of an older version is available on HuggingFace.

https://huggingface.co/spaces/TempoFunk/makeavid-sd-jax

#aiart #text2video #ai #generativeart

WebVid-10M

is a large-scale dataset of short videos with textual descriptions sourced from stock footage sites. The videos are diverse and rich in their content.

- 10.7M video-caption pairs.
- 52K total video hours.

https://maxbain.com/webvid-dataset

Train split:
http://www.robots.ox.ac.uk/~maxbain/webvid/results_10M_train.csv

Validation split:
http://www.robots.ox.ac.uk/~maxbain/webvid/results_10M_val.csv

Video data coming soon.

#openscience #dataset #ai

ImageNet

a dataset organized by WordNet. Each concept in WordNet, described by words or phrases, is called "synonym set" or "synset". With an average of 1000 images illustrating each synset, quality-controlled and human-annotated.

https://www.image-net.org

v2 - Winter 2021
https://academictorrents.com/details/8ec0d8df0fbb507594557bce993920442f4f6477
🧲
magnet:?xt=urn:btih:8ec0d8df0fbb507594557bce993920442f4f6477

v1 - Fall 2011
https://academictorrents.com/details/564a77c1e1119da199ff32622a1609431b9f1c47
🧲
magnet:?xt=urn:btih:564a77c1e1119da199ff32622a1609431b9f1c47

#openscience #dataset #ai

The Pile

a 825 GiB diverse, open source language modelling data set that consists of 22 smaller, high-quality datasets combined together.

https://pile.eleuther.ai

(DEAD) https://academictorrents.com/details/0d366035664fdf51cfbe9f733953ba325776e667
🧲(ALIVE)
magnet:?xt=urn:btih:0d366035664fdf51cfbe9f733953ba325776e667

#openscience #dataset #ai

Text To Video Model reusing Stable Diffusion as image backbone.

Online demo:
https://huggingface.co/spaces/TempoFunk/makeavid-sd-jax

#text2video #ai #stablediffusion #makeavideo