Wikidata has linked various concepts and lexemes to images. For lexemes and their senses there are not that many images yet. For concepts there are many, but many are proper nouns or complex concepts, e.g., "Berlin" or "Chile at the 1956 Winter Olympics"
For Wikidata lexemes a query with the Wikidata Query Service at https://query.wikidata.org may be ...
The trick is to use str inside the contains.
This query gives the expected result:
select ?item ?img
hint:Query hint:maxParallel 400 .
hint:Query hint:chunkSize 8000 .
?item wdt:P18 ?img .
filter contains(str(?img), "Montage")
Another way to achieve this is to set hint:Query hint:regexMatchNonString true and then use a normal regex.
Too long for a comment, but not an answer:
Ask your lawyer. (Or don't, depending on how serious your project is. No one is going to sue you about a hobby web page showing some logos.) Most logos are copyrighted and the owner has not given permission for such use, but depending on what jurisdiction you operate in, there might be exceptions such as fair use that allow you to use them without permission....
The GFZ Potsdam, a large research institute in Germany, has several groups working on hyperspectral data.
This page http://www.enmap.org/flights.html lists a table with several datasets from Europe, acquired 2011-2014.
Most of the DOI links on that page lead to the library of GFZ Potsdam, which offers a data download service. However, for some reason most ...
One source for the US (2018 election) is
The first of its kind, this collection consists of every campaign logo from the 2018 election for United States Congress. The archive is a tool to explore trends and typologies that reveal themselves only when viewed in aggregate.
(they offer a bulk download on demand, and I'll ...
The entire framework is outlined here
LLD - Large Logo Dataset - including training dataset downloads
Training Generative Adversarial Networks (GANs) for logo synthesis on such multi-modal data is not straightforward and results in mode collapse for some state-of-the-art methods. We propose the use of synthetic labels obtained through clustering to ...
Check out Old Maps Online: https://www.oldmapsonline.org/. They have an interactive map of the world, which you can use to locate georeferenced historic maps.
They have many more old maps that are waiting to be georeferenced. Anyone can volunteer to help georeference them.
An organization can bulk-upload un-geoereferenced maps to their collection.
Once a ...
There is a dataset from IBM IoT that has corrosion images, but I don't know how extensive.
I also found this paper, which cites that the images where given from an industry partner. You might try to contact ...
One potential dataset is the Selfie Data Set
Selfie dataset contains 46,836 selfie images annotated with 36 different attributes divided into several categories as follows.
Gender: is female.
Age: baby, child, teenager, youth, middle age, senior.
Race: white, black, asian.
Face shape: oval, round, heart.
Facial gestures: smiling, ...
Check out the ICDAR 2019 Robust Reading Challenge on Scanned Receipts OCR and Information Extraction: https://rrc.cvc.uab.es/?ch=13
You have to make a free account and then you can access the dataset, approximately 1000 images of scanned receipts.
For optical astronomy, SDSS is where you should be looking at.
ESO has digitized sky surveys (DSS) from the pre-digital era.
I have worked with astronomers for around 9yrs helping them with data science and none of them use stuff from really old catalogs.
Hope this helps.
Just released (11.2018)
https://fastmri.med.nyu.edu/ (access by request)
The anonymized imaging dataset provided by NYU Langone comprises raw k-space data from more than 1,500 fully sampled knee MRIs obtained on 3 and 1.5 Tesla magnets and DICOM images from 10,000 clinical knee MRIs also obtained at 3 or 1.5 Tesla. Curation of these datasets are part of ...