The Holy Grail of Coin Recognition – The BeSpecular App

When people are looking for coin identifier apps, often they end up on this page. This post is only slightly related, but I wrote a new one. Read about coin identifier apps here

Those of you who have been reading this blog for a while probably know that the biggest challenge of collecting without vision is recognizing an unknown coin. There just isn’t a good way to do it alone. I need help. I was trying to come up with all kinds of ways of using technology to do this independently, but I couldn’t succeed. There are a number of shortcuts, but in most cases I need help from a human. While this hasn’t changed, I found a solution where I don’t have to bug friends and family, and I can get a coin recognized any time I’d like to. I found an app, called BeSpecular. It is not a coin identifier app, but it is the closest I can get to one.

What is BeSpecular?

Bespecular is an iPhone or Android app, using crowd sourcing to help blind people identify anything they take a picture of. In a nutshell, I can take a picture, ask a question about it, send it to volunteers, and whoever is available and is willing, can describe the picture, either using text or voice, and I receive the response, usually within seconds or minutes. But much more about it later.

Why Bispecular?

There are many similar applications trying to achieve the same thing, approaching it from different angles. I won’t list them here, but they are all great. There are apps which use image recognition to analyze what is on the picture. In case of coins, the recognition is very poor, for most foreign coin all I get is “silver coin”. It may give me a little more info about US coins, but not enough. There are other services, where I can contact a real person using my phone’s camera, and ask for a description of anything. I tried those solutions as well, when I’m recognizing coins, the results really depend on how familiar people are with coins or what exactly I’m looking for on them. Of course, I can explain, but it can be very time consuming, and I found the results very minimal, as coin recognition requires somewhat of a specialized knowledge. I use these solutions for other things, such as setting the AC in a hotel room when I’m alone, matching shirts with pants, reading a screen when my screen reader freezes, etc. But BeSpecular takes a different angle. First, I post the picture, and only those people respond who exactly know what I’m looking for, and can read the coins. The pictures go to many volunteers, and only those people respond who can help with the answer, or interested in finding it out. Therefore, I get a much more accurate response.

How do I get coins posted for recognition?

It definitely took a while to figure out what works best. There are two components. Good pictures, and good questions. I’m stating the obvious here, but asking what kind of coin it is doesn’t get the right answer, at least not the one I’m looking for. But when I made the questions more specific, such as what is the country, value, currency and date, I started getting excellent responses.

The picture was more difficult, but I found a relatively easy way of taking a recognizable picture of coins. These are definitely not good quality, but it serves the purpose. I use a little phone stand to keep the phone levelled, approximately two inches above the coin. The coin is placed into a small plastic rectangle, I align it to the right edge, approximately the same distance from the three sides of the rectangle. This places the coin right under the phone’s camera.

How the process works?

Now comes the fun part. From the app, I take a picture of both sides of the coin, then I ask my question, and send it to the volunteers. This process takes about a half minute. I get a spoken notice when somebody starts to create a reply or when a reply arrives. I know exactly at any given time how many replies I have, and how many people are replying at the moment. So, as I go to read or listen to the first reply, usually a few more are coming. At any given time, I can say that I have enough replies and I don’t want to receive more.

But why would I receive more replies? Well, for many reasons. Primarily, because this is how the system works. But in case of the coins, different people give me different information. Sometimes I get all I need in the first reply. But there are instances when the photo isn’t too good, and people recognize different things on a coin, or while one person may say that they can only determine the value of the coin, another is more familiar with the coin, or the language, and provides me with other type of information. It happened several times that I pieced everything together from two or three partial descriptions. However, at other times, while the first reply is not sufficient, I choose to receive more, and several people send me good replies at the same time. At this point, somebody worked in vain. I usually try to stop the replies when I get one or two, and request more if I don’t have enough information.

Here is a screenshot of a reply where a volunteer, Sam, told me that it was a 10 Kronur from Iceland. I was slow to stop the replies, so after this I got two more. I’m still trying to get the hang of it while trying to be fair with the volunteers.

This service is too good to be true. I can pretty much any time have a coin recognized. Well, almost. Currently, each person is allowed six questions a day. Fair enough, some people might have more relevant questions, so why would volunteers recognize hundreds of coins for me? However, I hope there will be a paid option, because this is definitely something that’s worth supporting. I contacted the founder of the project, and she explained that they are working on other options.

I hope this service will receive funding and recognition and will stay alive, it definitely has a room in the blind person’s toolkit. There are times when automatic image recognition is just not good enough, and getting on a video call with somebody is not an option, or just not necessary. Personally, I frequently use all three approaches.