cc-ai / kdb Goto Github PK

People like Karthik (Climate Science people probably) are not used to Git/Github -> we should put together a quick description/documentation + ressources to get them started asap with our workflow on github

To hold the executive summary presentations and so on

@ClementFeltin is part of an NGO creating CC educational resources and they'd like to be part of a potential beta-testing programme we may have some day

Devon has started working on this

Add to workflow description

Follow up pour savoir si Linda a fait le taf

see https://github.com/cc-ai/floods-frontend

@ClementFeltin do you think it may be possible to meet the cognitive scientists you know, sometime in June?

The website for the crowd source image collection project (we need a better name by the way) needs some actual UI design. I have create some basic prototypes just to have something from which to work from but in the grand scheme of things, if we want to use this project for more than floods, we will need to think about this. Here is a list of the main functionality I think the website should possess:

• Usable through a web page or a mobile app. We will use React Native for this. I have the Android part covered, however, even though React Native create the necessary files to build a project for iPhone, we still need a Mac, a XCode license, an Apple Developer license and to tailor the app for the AppStore guidelines.
• User can upload one or more picture(s) from their phone or take one using their camera.
• User can read about who we are, the goals of the project and legal stuff for their images.
• Once a user has uploaded their image(s), they can preview what they will send and delete those they do not want to send anymore.
• The user can send their image(s) or cancel.

For me, this is the bare minimum. Now, if we want to scale this to other type of natural disasters, my concerns are as follows:

• We can either leave the website as is. This will allow us to quickly promote it and hopefully reach an audience. However, we will need to skim through all the pictures ourselves to organize them by natural disaster.
• We can add a description box. But then the user needs to enter text and the process is slower and more tedious which might make them less inclined to use it.

Edit:
Suggestions to be RGPD compliant:

• Handling creation of an account.
• Provide the user with a unique ID associated with the images. They can then contact us by email to get their images removed.

If we split the generative work into generating water, generating sky etc. (SPADE-like)

To smooth things up even further we need an issue template.

Ideally with the project attribute pre-set. Suggestions?

reference sections in root readme.md

Find or design economic models which can convert actions into climate variables

We need to work on this workflow so everyone should start adding issues, even if they are not critical, just to get a feel of how things work issues, labels, assignments, interaction org <-> project <-> repos, the role of kdb and so on

We could have a file where everyone would be welcome to share their experience / background / interests / affiliations / ..

For instance: (Contribution = domains you intend to get involved with)

(columns are debattable)

So we know what papers we've had a look at

• formulate needs if we get in touch with Google

Given we have chosen the right emotions

Check out https://github.com/cc-ai/kdb/tree/master/workflow#good-practices and speak you mind :)

@DatCorno could you add this repo's labels to yours?

-> follow instructions at https://github.com/cc-ai/kdb/tree/master/workflow#labels

-> give me feedback on clarity + ease of use / propose improvements

Thanks!

in ressources/code/ and ressources/papers/

One of our solutions to data-scarcity is to use simulations to generate floods for instance. Any help wit that (GTA-V especially) would be welcome

problem 1

including expected implementations

Send individual notifications for mentions/assignments and so on

From a segmentation of a non-flooded house, generate the segmentation of the same house

• learn from simulation data
• geometric approach if depth map available (water plane)

Most of them are already there, add the one about issues

Figure out suitable model within the region compared to other flooding models elsewhere.

Project abstract:
https://docs.google.com/document/d/1jhmWQGgHQ__kid-D-JMR_V5Ao41t-OXKyrplvrY8pcc/edit

->

Share backend API key (privately) w/ @elenab

We'll soon get the code we might have to plug it into LabelBox

example issue for tutorial

We're currently using https://www.labelbox.com/ to label/segment our images.

And I'll soon talk with the guys from CurveGCN.

Get in touch if you have some spare time :)

Waiting for Arxiv upload

Can MSR help with the cc-ai organization?

Conditioned on the data we get