The Watchmakers App

Posts
46
Likes
15
The new iOS app for watch enthusiasts just got released!

The app helps easily identify watch movements caliber (with most advanced filtering options), and also the most up-to-date list of U.S. import codes found on the movements. The app and the data is being constantly improved and updated.


Hope you will find it useful. Get it at thewatchmakersapp.com!
 
Posts
2,444
Likes
3,898
Looks like AI based. This also looks like an unsolicited sales post.

Would be nice if someone did an AI of the bestfit catalog.

The data is there. Most of this sort of thing is done with the least amount of effort where they expect maximum return.

Edit: I clicked on the link so you do not have to.

Screenshots make it look like it is a scrape of the newer Ranfft site. Not sure where the data came from but it could violate someones IP as this information was only collected in a few places. Technically one of the materials houses re copyrighted the bestfit catalog.

The real question is who is checking and vetting this data. There is a lot of misinformation out there. Scrapes of online data are not the best sources for this sort of thing.

I only mention this as I have been thinking about this sort of thing for 30 years. Of course AI renders much of this moot, at least in the short term becouse, man eating chickens often frequent the midway, and no one wants to admit that they took time and paid to see the goldbrick.
Edited:
 
Posts
46
Likes
15
Looks like AI based. This also looks like an unsolicited sales post.

Would be nice if someone did an AI of the bestfit catalog.

The data is there. Most of this sort of thing is done with the least amount of effort where they expect maximum return.

Edit: I clicked on the link so you do not have to.

Screenshots make it look like it is a scrape of the newer Ranfft site. Not sure where the data came from but it could violate someones IP as this information was only collected in a few places. Technically one of the materials houses re copyrighted the bestfit catalog.

The real question is who is checking and vetting this data. There is a lot of misinformation out there. Scrapes of online data are not the best sources for this sort of thing.

I only mention this as I have been thinking about this sort of thing for 30 years. Of course AI renders much of this moot, at least in the short term becouse, man eating chickens often frequent the midway, and no one wants to admit that they took time and paid to see the goldbrick.
To address your concerns, it is not a scrape of the site mentioned by you (there is a lot of misinformation on that one) and data is being checked/ vetted/updated. There are gaps and omissions, but those are being filled it. Even at this moment this might be the most precise database around.

Tip on how to quickly check the quality of database: search for Valjoux 55 and check if it has 17 jewels.
 
Posts
21,302
Likes
48,472
It looks like chatgpt can identify watch movements from photos.

Sometimes ... but it also hallucinates. ::facepalm1::
Edited:
 
Posts
46
Likes
15
It looks like chatgpt can identify watch movements from photos.

Sometimes ... but it also hallucinates. ::facepalm1::
Identify from photos - that's very unreliable. There are too many movements looking the same, as there are many bridge variations and other variations of the same movement. Or consider automatic movements - rotor position can be different to compare with the photo.

The problem with AI and ML is that there is no reliable dataset to train it on.
 
Posts
21,302
Likes
48,472
D deni2s
Identify from photos - that's very unreliable. There are too many movements looking the same, as there are many bridge variations of the same movement.
With a good enough training set, it would be reliable. Clearly LLMs are not the best type of AI for this, but conceptually it's a trivial AI application. The problem is that nobody is financially motivated to assemble the training data.
Edited:
 
Posts
27,747
Likes
70,501
D deni2s
Identify from photos - that's very unreliable. There are too many movements looking the same, as there are many bridge variations and other variations of the same movement. Or consider automatic movements - rotor position can be different to compare with the photo.

The problem with AI and ML is that there is no reliable dataset to train it on.
What is your involvement in this?
 
Posts
46
Likes
15
With a good enough training set, it would be reliable. Clearly LLMs are not the best type of AI for this, but conceptually it's a trivial AI application. The problem is that nobody is financially motivated to assemble the training data.
Challenging, but I am assembling it. What's currently in the app - those are just first steps. More to come.
 
Posts
46
Likes
15
What is your involvement in this?
I have done some experiments with trying to identifying the movement by the photo.
 
Posts
27,747
Likes
70,501
D deni2s
Challenging, but I am assembling it. What's currently in the app - those are just first steps. More to come.
So this is your app?
 
Posts
6,499
Likes
26,440
So this is your app?
Given that his username resembles the name Deniss, I would wager that it is

 
Posts
701
Likes
2,601
If deni2s is Deniss, he is also the developer of the Ranfft app.
 
Posts
2,444
Likes
3,898
The "in App purchases." is the tell here.

As noted the training data is not difficult to come by. Most manufactures published parts lists showing the parts with 1:1 scale. Fax and JPG does mess this up. Some even complain about Bayer mosaic. I once worked back in the 1990s for a company that was attempting to de halftone images. All these add high frequency structured noise to the image.

My guess it Chat GPT and the offspring probably already scraped Ranfft and the online datasheets, which are usually PDF copies of faxes.

Most real watchmakers have these reference datasheets. Or a copy of the Bestfit catalog which did aggregate the manufactures data sheets. These were designed so that one could use an optical comparitor to identify parts by outline. The Bestfit catalog has the calibers sorted by ligne size and outline. This is fairly trivial. Chances are like omega, this data was stored on punched cards use by tabulating machines. Omega did.

The parts "inventory" designations are a standard by the manufactures. So part 721 will always be a balance wheel. Some like Omega had their own IBM type mainframe, so they used 1327 for a balance complete. There are cross conversion charts that the trade uses.


Much of this was scrapped, trashed in the shredder or furnace. What remains is considered of value to the manufactures, so the release of this data is probably not likely in this world, which like the old one values information as power. This is the manufactures IP.

Dr Ranfft did have a nice parametric search algorithm. This does not work on the archive.org data. The new scraped site (which this seems to be.) does not understand this or how it worked. It really is about the subtlety of different bridge layouts.

As for specific calibers. In the last week I have had to deal with A Schild 1895/1896 which shares a lot of parts with other calibers. There is also the thread about the Venus 206. So there are going to be corner cases.

That is not to say there is not value in an app such as the OP seems to have written. It probably though is not an area what can be monetized easy. Hobbyist ,You tubers, and fans might find something like this of value. There is 300+ years of data out there, so it is nice to learn exatly what that box of old Swiss cylinder movements really is.

I scanned 100s if not 1000s of piano rolls, which were converted to MIDI. There were 100 people on the mailing list 25 years ago. Of these probably half built scanners and scanned the same group of 10,000 piano rolls. (I focused on pipe organ playing rolls.) So much of the same observations apply. One can not actually sell or share the data since there are 'owners.' who use this data as collateral for securing loans. It remains a hobby activity.
 
Posts
27,747
Likes
70,501
D deni2s
Yes
Okay, that explains the posts here and on WUS promoting it. People would likely be more accepting if you were up front about this being yours…
 
Posts
46
Likes
15
Dr Ranfft did have a nice parametric search algorithm. This does not work on the archive.org data. The new scraped site (which this seems to be.) does not understand this or how it worked. It really is about the subtlety of different bridge layouts.
I am not sure if you are actually talking about The Watchmakers App. It already contains more than 100 working parameters for searching/filtering:
  • Manufacturers country
  • Manufacturer name
  • Power source
  • Setting type
  • Complications
  • Dial configurations
  • Hands configurations
  • Movement size
  • Movement shape
  • Jewels count
  • Timepiece designation
  • Pocket watch specific features
  • Escapement type
  • and many other parameters like Geneva stopworks, for example, and so on.

And there will be more added in future. They serve their purpose and in my experience they do it well.

BestFit is good, but it's still not ideal - and yes, it is compilation of other sources. There are many other catalogues, old books, technical sheets, etc. containing more info. If I need to identify some movement caliber I might need to search for it in 5-10 different catalogues, each of them having different structure, different terminology, etc... It takes time to do that. So you can spend an hour looking into catalogues, or you can match it in 5 minutes using the app. And you probably know the prices of those catalogues/books, especially the vintage ones, some of which are very rare. I have already invested crazy money in many of those to improve the data. So with this app you can save money and what's more expensive - time.

The current information about the calibers in the app is available for free. The advanced search is under subscription with 2 weeks trial period and you can cancel it in any time. It was not an easy walk in the park to structure the data and make the search reliable and fast.

And this is just the first version of the app, it's not that it's finalised - I am thinking currently it is in MVP (Minimum Viable Product) phase, so there are more features and information to come in future versions.

About specific calibers and corner cases... I am aware of that. I am vetting the data, I am taking notes, I am refactoring and updating the data in the database. I have a long list of missing calibers. I know that people want to see lift angles, technical sheets and many other things in the app. They will probably come. And probably more features will come, which haven't even imagined. But I am also taking systematic approach, I have my roadmap to make sure no important data is lost or messed up (like some scrapers did) and that it is reliable and solid base for future features.

If you notice any omissions, typos or other errors in existing caliber descriptions - please let me know. I have already fixed a huge amount of information (that's why I am so sure that currently it is the largest most up-to-date movements database available), but with large data like this it's not so easy to notice and fix everything. If there is someone knowledgable in some subset of movements (be it some brand or type) and wants to systematically improve information, so it is available for other enthusiasts - please contact me.

I believe this app will be a game changer. And there is at least one very interested user for it - me 😀
 
Posts
46
Likes
15
Okay, that explains the posts here and on WUS promoting it. People would likely be more accepting if you were up front about this being yours…
Don't kill the messenger! 😁

I am interested in non-subjective feedback and constructive criticism. It's totally fine and understandable that people are very suspicious when looking at what looks like "yet another scraped data clone generated at night using AI trying to get easy money and fame".

But yes, I decided to use the release as a good excuse to promote it a bit too, to get some early adopters and early feedback from actual users. Having some actual users is a great motivation.
 
Posts
2,444
Likes
3,898
D deni2s
Don't kill the messenger!
Funny how people respond to advice and constructive criticism, when they really want attention and admiration. Not to mention platitudes.

If one reads the room here, this is not the place for self promotion.

Most of us who have been collecting for decades already have the reference books and data sheets.

The search terms mentioned are all marketing fluff and of minimal use. Especially as this is designed to monetize and profit from the work of others.

To be useful the search needs to be parametric and graphical. Ideally able to take a photo and identify a part. Yes lift angles are useful, but there are easy to find charts of these.

here is a photo of some parts I got online recently:



This is how many parts were packaged.

Note that the circle is the shape of the movement 1:1. Which in this case is round. The numbers are measurements. The silhouette shows the setworks, which are the fingerprint of the base caliber. The name of the caliber is below.

Now what is in the box is something else. In this case a bunch of hands and screws. These may or may not be from Enicar watches.



I offer these observations as I do feel there is a good need for some modern tools for parts searches. Better yet would be to provide a secure insured reliable fast shipping service.

Material houses are full of boxes like the one above. Stocks of spare parts what are enough to cover hundreds of years of service and support. Of course the popular stuff is picked over.


There is already a good foundation in the old hard to use thing like the bestfit catalog. All I am suggesting is to buld upon it. The data is there for someone to parametize it. Much of it is in easy machine readable format. Byproduct of a day where index cards and phone directories were the main way databases were accessed.

What is really needed is an app what can search material houses. Which like OFrei have horrible search results. Paylak (watchmaterial.com) is almost as bad. Always returning the same 1700 results for Omega searches. There is no way to do sub searches. I have a window with 4 tabs in the background with the 4 or 5 major material houses open. Usually the only way to find a part is to call them with specifics. And even then they do not like going into the dusty stacks and look through 1000s of drawers. Some even request you send the inquiry by email which is really easy to ignore.

When such places fail, the stuff is either dumped on ebay, or simply carted off to the landfill or metal recycler.

My mentor said post 1940s watches could last 500 years or more. I am reading a book Revolution in time by David Landis. I have not read this in 25 or 30 years, but some of the subject has stayed with me. This is a book on economics. A bit depressing as it is about companies, countries and other things what have failed after dozens if not hundreds of years of success. The book was written in the 1980s during the quartz crisis. Probably to address some of the economic failings happening in places like Switzerland. A lot of the ideas now seem dated.

Watchmaking is about high volume. Since the 17th century watches have been produced in the 10,s 100s and millions of units per year. The book seems to imply that mechanical watches will be of no value since quartz was better. Yet the numbers of mechanical watches produced remained high. Here 40 years later, mechanical watches have the same appeal they did in the 18th century, and for the same reasons.

10,000 (why is it always 10,000) is a mere drop in the bucket of what has been made in the last 3 or 400 years. And cell phones are pocket watches providing the same functionality.

99 or more percent of people could not care about all the defunct and failed brands. The brands what do have 'history.' go to great lengths to obfuscate the real details. Vis. the Longines thread here? Did the OP even read that. Comment on it. Ask questions? they even have not taken the time to create an avatar.

Yes the app might be useful, to the same person who likes watching You tube tear down videos. On the other hand as soon as subscriptions or free trials become involved. Then we are in the same place as a carnival clip show. Which is no different than a man eating chicken. Which was where the term Geek comes from. This was the person who bit the heads off live chickens. Except it was a stage illusion. A magicians misdirect.

Currently dive watches are popular. Chronographs still retain some interest. Trends shift. Between the 1930s and the 1970s it was illegal for US citizens to own gold specie. So the value was put into things like watch cases. I had to look up the word specie. Landes uses that word a lot in trade deals. The main reason people will search for obscure watches is to see if it is the watch which is equivelent to winning a lottery. Ironically Landes says to get around trade restrictions many watch companies turned to subscription draw lots to shift excess inventory.

I also think in this day and age of constant tracking. Users want anominity. The ability to not have to 'like and subscribe' or create yet another set of login tracking credentials.

This is intended as Constructive critsism. Tough love in a way. There is no easy way to any of this, or else it would be popular and many people would be doing it.
 
Posts
46
Likes
15
Most of us who have been collecting for decades already have the reference books and data sheets.
Many still don't have and I see a lot of heresy shared in the forums. And then there are people who just started and can't find info anywhere.
The search terms mentioned are all marketing fluff and of minimal use.
You are quick with your conclusions. I tend to disagree, I was able to filter out and identify many movements.
Especially as this is designed to monetize and profit from the work of others.
I think you got this totally wrong. The "work of others" I made available for free to anyone a year ago by releasing Ranfft app and I supported it, so that people can have that info which otherwise would be lost (good luck relying on data from ranfft.org or emmywatch!). And I spent my time and efforts and there is no monetisation at all. And maybe that's the reason I didn't had time to choose some not meaning anything avatar on every watch forum I am on.

And even in this new app, all that information is pretty much improved/organized/structured, and it's still available for free, not behind the paywall. Yes, I am trying to monetise additional features (like search filters at the moment), but that's where I put myself 100% effort.

I am still wondering if you actually checked the app, as you seem to be very quick at drawing some conclusions based on first impressions like avatar image...
To be useful the search needs to be parametric and graphical. Ideally able to take a photo and identify a part. Yes lift angles are useful, but there are easy to find charts of these.
The search is parametric unless we have different understanding about what it means. Again check the app please.... About identifying the part from the photo - let's return back to reality 😉
I offer these observations as I do feel there is a good need for some modern tools for parts searches. Better yet would be to provide a secure insured reliable fast shipping service.
Great idea, I totally support it! Eager to check your implementation when it's done 😉
There is already a good foundation in the old hard to use thing like the bestfit catalog. All I am suggesting is to buld upon it. The data is there for someone to parametize it. Much of it is in easy machine readable format. Byproduct of a day where index cards and phone directories were the main way databases were accessed.
I can't imagine how much work, resources and effort was spent on creating something like Bestfit in those days when there was no internet available... I plan to rely first on technical sheets and catalogues from manufacturers as foundation and after that maybe will double check with BestFit (also considering that there is some copyrights holder for Bestfit).
What is really needed is an app what can search material houses. Which like OFrei have horrible search results. Paylak (watchmaterial.com) is almost as bad. Always returning the same 1700 results for Omega searches. There is no way to do sub searches. I have a window with 4 tabs in the background with the 4 or 5 major material houses open. Usually the only way to find a part is to call them with specifics. And even then they do not like going into the dusty stacks and look through 1000s of drawers. Some even request you send the inquiry by email which is really easy to ignore.

When such places fail, the stuff is either dumped on ebay, or simply carted off to the landfill or metal recycler.
Sounds like material houses are quite happy with their current business model, otherwise they would change something, don't you think? They have enough resources to make it happen easily if they would want it.
Here 40 years later, mechanical watches have the same appeal they did in the 18th century, and for the same reasons.
I wouldn't say the same... The difference being that in 18th century not many could afford the watch, when now not that many want the watch.
10,000 (why is it always 10,000) is a mere drop in the bucket of what has been made in the last 3 or 400 years. And cell phones are pocket watches providing the same functionality.
Ok, currently it's exactly 10,944 calibers. And probably that covers like 50%-70% of movements (not calibers) ever made. Most of the data is from 1940-1970 era. I plan to expand that range.
99 or more percent of people could not care about all the defunct and failed brands. The brands what do have 'history.' go to great lengths to obfuscate the real details. Vis. the Longines thread here? Did the OP even read that. Comment on it. Ask questions? they even have not taken the time to create an avatar.
Probably not. Personally currently have more interest in Zenith watches. Have to prioritise, don't have several lifetimes to read and comment everything.
Yes the app might be useful, to the same person who likes watching You tube tear down videos. On the other hand as soon as subscriptions or free trials become involved. Then we are in the same place as a carnival clip show. Which is no different than a man eating chicken. Which was where the term Geek comes from. This was the person who bit the heads off live chickens. Except it was a stage illusion. A magicians misdirect.
Thanks for supporting! 😁
I also think in this day and age of constant tracking. Users want anominity. The ability to not have to 'like and subscribe' or create yet another set of login tracking credentials.
Everything has it's price. There are some strange people around who even register on forums and share photos there... Main intention why sign in is there in the app is not to track, but to provide better user experience considering the planned features.
This is intended as Constructive critsism. Tough love in a way. There is no easy way to any of this, or else it would be popular and many people would be doing it.
Not that I agree with everything you wrote, but thanks anyway for your opinion, I will keep it in mind. Wanted to pick the avatar, but spent all the time replying... 😁
 
Posts
2,444
Likes
3,898
Thank you for taking the time to answer my questions.

I think you answered points to my satisfaction, and probably others here as well.

Your primary goal is not to engage and participate this community. Rather it is to sell something ( a service) and direct people to your website and download subscriptions apps.

That you admit to joining many forums and not take the time to create an avatar, speaks to the point. This is against much ot the spirit of things here. And why there is a 200 minimum post before soliciting sales.

I am sure that people can find your aggregator useful. I find that Archive.org serves much the same purpose. And yes I did not download the app. For the reason mentioned. It says that there are in app purchases. No it is not a paywall. But the effect is the same.

Some of the more interesting things, which I find stimulating, are dismissed out of hand as being too difficult. Mostly though it is time consuming. Which at this point is not economical.