Embed Notice

HTML Code

<blockquote style="position: relative; padding-left: 55px;"><section><a href="https://mastodon.nz/users/weiji14/statuses/113629851479266948">偉吉 (weiji14@mastodon.nz)'s status on Wednesday, 11-Dec-2024 04:15:48 JST</a><a href="https://mastodon.nz/@weiji14" title="weiji14@mastodon.nz"><img src="https://gnusocial.jp/avatar/305143-48-20241210191548.webp" width="48" height="48" alt="偉吉" style="position: absolute; left: 0; top: 0;">偉吉</a><div><ul><li><li><a href="https://gnusocial.jp/user/249175" title="excalidraw@mastodon.social">Excalidraw</a></li></ul></div></section><article><p>Presenting my AGU24 poster this Tuesday afternoon - GC23A-0229: Choosing Earth Foundation Models: From research grade to community maintained. Come have a chat about how we can make Foundation Models for Earth Observation / Geospatial data more practical for different use-cases!</p><p>Poster made using <a href="https://mastodon.social/@excalidraw">@excalidraw</a>, code available at <a href="https://gist.github.com/weiji14/7f3ade1ad231587380f131c402b9649d" rel="nofollow noreferrer">https://gist.github.com/weiji14/7f3ade1ad231587380f131c402b9649d</a>.</p><p><a href="https://mastodon.nz/tags/AGU24" rel="tag">#AGU24</a> <a href="https://fosstodon.org/@developmentseed">@developmentseed</a> <a href="https://mastodon.nz/tags/EarthObservation" rel="tag">#EarthObservation</a> <a href="https://mastodon.nz/tags/MachineLearning" rel="tag">#MachineLearning</a> <a href="https://mastodon.nz/tags/BetterPoster" rel="tag">#BetterPoster</a></p></article><footer><a rel="bookmark" href="https://gnusocial.jp/conversation/4160872#notice-8132627">In conversation</a><time datetime="2024-12-11T04:15:48+09:00" title="Wednesday, 11-Dec-2024 04:15:48 JST">about 6 months ago</time> <span>from <span><a href="https://mastodon.nz/@weiji14/113629851479266948" rel="external" title="Sent from mastodon.nz via ActivityPub">mastodon.nz</a></span></span><a href="https://mastodon.nz/@weiji14/113629851479266948">permalink</a><h4>Attachments</h4><ol><li><label><a rel="external" href="https://gnusocial.jp/attachment/3672400">Poster titled Choosing Earth Foundation Models: From research grade to community maintained. Center part of poster are the names of 7 models (ScaleMAE, SatMAE, Satlas, Prithvi-EO-2.0, Presto, Clay, DOFA), each with bar charts of different lengths extending out in the left, bottom, right and top direction, indicating varying 'levels' of modalities, computational complexity, documentation and licensing. The main text for the 1st panel reads:0) Why Earth Foundation Models?Earth Foundation Models offer a way to get a headstart
on building models for analyzing Earth features. While
the initial self-supervised learning pre-training phase
is expensive, it can be done once by a few players, and
then be adapted for multiple downstream applications.To maximize the use of these Foundation Models,
we argue that it is not sufficient to just architect
models with fancy technical features. More attention
needs to be given to empower community participation,
and ensure that long-term maintainability is prioritized,
so that Foundation Models can be relied upon as a tool
for tackling complex environmental challenges that lie ahead.</a></label><br><a href="https://mastodon.nz/system/media_attachments/files/113/629/805/498/294/585/original/d7785b0a79726ca8.png" rel="external">https://mastodon.nz/system/media_attachments/files/113/629/805/498/294/585/original/d7785b0a79726ca8.png</a></li><li><label><a rel="external" href="https://gnusocial.jp/attachment/3672401">Untitled attachment</a></label><br></li></ol></footer></blockquote>

Corresponding Notice

Embed this notice
偉吉 (weiji14@mastodon.nz)'s status on Wednesday, 11-Dec-2024 04:15:48 JST偉吉
- Development Seed
- Excalidraw
Presenting my AGU24 poster this Tuesday afternoon - GC23A-0229: Choosing Earth Foundation Models: From research grade to community maintained. Come have a chat about how we can make Foundation Models for Earth Observation / Geospatial data more practical for different use-cases!
Poster made using @excalidraw, code available at https://gist.github.com/weiji14/7f3ade1ad231587380f131c402b9649d.
#AGU24 @developmentseed #EarthObservation #MachineLearning #BetterPoster
In conversationabout 6 months ago from mastodon.nzpermalink
Attachments
1. Poster titled Choosing Earth Foundation Models: From research grade to community maintained. Center part of poster are the names of 7 models (ScaleMAE, SatMAE, Satlas, Prithvi-EO-2.0, Presto, Clay, DOFA), each with bar charts of different lengths extending out in the left, bottom, right and top direction, indicating varying 'levels' of modalities, computational complexity, documentation and licensing. The main text for the 1st panel reads: 0) Why Earth Foundation Models? Earth Foundation Models offer a way to get a headstart on building models for analyzing Earth features. While the initial self-supervised learning pre-training phase is expensive, it can be done once by a few players, and then be adapted for multiple downstream applications. To maximize the use of these Foundation Models, we argue that it is not sufficient to just architect models with fancy technical features. More attention needs to be given to empower community participation, and ensure that long-term maintainability is prioritized, so that Foundation Models can be relied upon as a tool for tackling complex environmental challenges that lie ahead.
  https://mastodon.nz/system/media_attachments/files/113/629/805/498/294/585/original/d7785b0a79726ca8.png
2. Untitled attachment

Public

Embed Notice

HTML Code

Corresponding Notice