Commons:Free media resources/Photography/Crowdsource app

MV Likoni, uploaded via Crowdsource, and best picture of this ship we have so far.

We had "Textile Mills in ..." categories for most US states, but zero picture for Pakistan, the 4th largest textile producer in the world.

The Crowdsource app markets itself as a way to fight bias in Artificial Intelligence. Thanks to the developer's strategy/campaigns, users are mostly people from digitally underrepresented countries. Users can submit pictures of their surroundings, and are asked whether they want to make their contributions CC-BY or not. About 400,000 CC-BY pictures have been uploaded so far.

How to import

1. If you have a Google account, go to https://crowdsource.google.com/images.
2. In the search bar, type something like locomotive/street/sign/device/etc or a country/city/brand name.
3. In the search results, choose any picture that:
   • Has encyclopedic value.
   • Embeds enough context in the picture itself. For instance, if it is unclear in what country a nature picture has been taken, then it is not useful.
   • Complies with all Commons policies. For instance, remember that many African countries do not have freedom of panorama.
   • Has "CC BY 4.0" written at the bottom of the webpage.
4. Right-click to download the picture.
5. Using the Upload Wizard, upload the downloaded file and information, in particular and picture URL and author name which caan be found at the bottom of the webpage.
6. In addition to other relevant categories, add this category to the picture: Contributed via Google Crowdsource app
7. The Upload Wizard forces you to select a precise date, so choose today's date then edit the uploaded picture's page to replace it with: {{other date|between|2016|2022}}

Alternatively, for instance if you do not want to use a Google account, pictures and metadata can be downloaded in bulk.

Why this dataset can be beneficial to Commons and Wikimedia

• Mostly from Africa and other underrepresented regions. These are the regions from where Commons and the Wikipedias need pictures the most.
• All pictures are recent.
• About a third of the pictures are of good photographic quality.
• A good 5% of the pictures show something that has a Wikipedia article but no Commons picture yet.

Warnings

• The pictures have no geolocation nor description. This limits us to pictures where the subject can be identified with 100% certainty, for instance helicopters/superyachts/places/etc.
• Labels shown on the side of each picture are unreliable. At best they can be a starting point for investigation, for instance if a picture of a stadium is labeled "Monrovia" you may want to search for stadiums in Monrovia and see whether any match the one in the picture.
• Pictures are not dated. We only know that all pictures were uploaded between 2016 and now.
• All faces are blurred. This is not a good dataset if you are looking for faces of people.
• The web search interface only shows the first 50 results, so be creative with search keywords, or download in bulk and write a script.