[OSM-talk] Pictures of opening hours signs for machine learning purposes

[Bryce Cogswell]

Here’s my first stab at automatically generating opening hours strings from photos provided by Lukas. There isn’t any attempt to make the output pretty or to rationalize it, I’m just detecting days/times and trying to form a valid opening_hours string:

http://gomaposm.com/opening_hours/index.php

While the results are mixed I feel that real-world results would be better for three reasons:
* The photos here were taken at off-angles for privacy reasons, which wouldn’t need to be the case when part of a phone app
* When translating with a “live” camera feed the user can move the phone until the translation is correct
* These results were done on my Mac, but ironically the OCR works better on the iPhone than the Mac (maybe because my phone has the iOS 14.5 beta running on it). 

Bryce

> On Apr 6, 2021, at 1:13 AM, Isaac Boates <iboates at gmail.com> wrote:
> 
> I would be interested in trying something out as well, Amazon web services has some very powerful pre-built image text recognition services as well, which are very easy to use. I have never tried it against hand-written text but would be interested in seeing how it fares.
> 
> I will also send you a private upload location.
> 
> Isaac
> 
> On Mon, Apr 5, 2021 at 9:37 PM Bryce Cogswell via talk <talk at openstreetmap.org <mailto:talk at openstreetmap.org>> wrote:
> > Great to see that someone is interested! How shall we proceed? Can you provide some kind of online storage where I can upload the files? Would you like to start with only a subset so that you can check on feasibility?
> 
> I’ll send you a private message with an upload location (it’s probably easiest if you upload everything). If other people are interested please let me know and I’ll provide a link for download. 
> 
> Bryce
> 
> 
> _______________________________________________
> talk mailing list
> talk at openstreetmap.org <mailto:talk at openstreetmap.org>
> https://lists.openstreetmap.org/listinfo/talk <https://lists.openstreetmap.org/listinfo/talk>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk/attachments/20210408/8fc7ab28/attachment-0001.htm>