Recent searches
Search options
Run Linux, post on the fediverse, and care about accessibility?
You need Frog.
https://github.com/TenderOwl/Frog#readme
Frog extracts text from images.
Use it to copy and paste the text in images into your image descriptions and help make the fediverse more inclusive.
@aral
So Frog should actually be an integrated part of the Fediverse, either at server or client level. As soon as I upload an image with text, Frog should run and add alt text automagically.
@joenepraat @aral Whoa, that's nifty. Didn't know that. And in that case, it should pop up and ask if you want the image to be OCR'ed automagically.
@joenepraat @aral tried the mastodon thing now, and it works really great. It's not 100% perfect with letters outside the english alphabet (at least not norwegian æ ø å), but it's super helpful to get the most of the work done quickly.
@joenepraat @cmyrland @aral yeah, it just never worked for me…
@cmyrland It’s not really a replacement for a human being explaining what is in the image to another human being. But it is a useful tool when your image contains the text.
@aral Yes, of course. I don't expect a relatively simple program to be able to describe a picture - not even the most powerful AIs are capable of that yet.
But for OCR-ing text it's very helpful, especially as a screenshot from a webpage can contain lots of text, and you don't always know the source for simple copy/pasting..
@aral thanks it looks really good. I'll try to use it
@tio Looks very neat; thanks :)
@aral
With the rate of current software development, it probably won't be too long before there's a free to use (but sadly likely not open source) online tool which examines an image (and optionally the metadata, e.g. Geotagging) and suggests various lengths / details of descriptions (e.g. depending on where you intend to post the image and the context of accompanying text, people may need to know it's a hanging basket, a hanging basket with red, yellow and blue flowers, or a hanging basket with red petunias, yellow impatiens and blue lobelia, or a hanging basket on Felpersham High Street etc etc). I'd put my nonexistent bets on a certain company famous for its search engine but less famous for being the world's largest online ad broker being (one of?) the first.
If the developers were well resourced enough to offer a free API, you could potentially see FOSS tools interfacing with it (likely offering to resize and/or strip all EXIF data - if the online tool could generate the description of what's in the image, the FOSS tool could add in geolocation in a privacy respecting way (e.g. Hooking in to any locally stored OSM maps)).