About this site
HANBAIKI.FYI is part documentary project, part educational exercise. The idea to record and map some of the thousands of vending machines (自動販売機・jidou hanbaiki) in Japan was there for a while. However, there is no way a single person could ever get around to do this on their own. So, I decided to crowdsource it. At the same time, I wanted to toy around with some new technology, including image recognition, that would help with sanitization of people's submissions.
Of course, this site may well remain a personal side project. However, I would be happy if some people do find value in the information, or better still contribute some photos of their own.
How can I upload new photos?
To contribute photos, please create an account first. Once you are logged in, you will see an Upload button in the navigation bar.
What happens to photos uploaded to this site?
At a high level, uploaded photos are processed as follows:
- Analysis: We read EXIF data embedded into the image by the camera to extract the GPS coordinates of the subject. If those GPS coordinates could be read successfully, processing continues.
- Recognition: We use image recognition to check if uploaded files indeed contain vending machines. If the confidence score for an image lies above a certain threshold, processing continues.
- Tagging: We use text recognition to check if uploaded files contain any known Tags (e.g., beverage brands). If the confidence score for a given tag lies above a certain threshold, the tags are automatically associated to the uploaded vending machine.
- Publication: Processed images are published on our site alongside their GPS coordinates and basic location information, such as state (県・都), city (区・市), postcode (〒), and town (町・村), if available.
Why can I not take photos directly from the upload form?
This site relies on the EXIF data embedded within the uploaded files to determine a photo's location. Unfortunately, the default camera dialog on iOS and Android -unless accessed directly via your camera app- does not embed GPS coordinates. Therefore, it is only possible for us to process photos that are already taken and saved, for example, in your gallery.
Why should I trust this site?
Luckily, you don't have to. The entire source code is open-source and available on GitHub.
I found a bug! What should I do now?
Please take the time to file an issue on GitHub.