Adrian Lopez

  • Content Count

  • Joined

  • Last visited

About Adrian Lopez

  • Rank

Contact Methods

  • Website URL

Recent Profile Visitors

349 profile views
  1. That's what I figure. I'm guessing a lot of those books haven't moved in a while because of their high prices and limited audience. An auction format will get things moving: you either bid on it now or risk missing out if it sells. Good for Tarisio, I suppose, but not so much for those who miss out on the auctions for not being ready to pull the trigger.
  2. It looks like the Tarisio bookshop is no more. The button for it is gone from the website's menu and leads to a "page not found." I had my eye on some books I was hoping to buy once finances improved, and while it's still possible to find them elsewhere the Tarisio bookshop was a convenient place to look for books on violinmaking. It's too bad, assuming it's not a glitch.
  3. According to their website, The Strad's online archive only goes back to January 2010.
  4. I don't know what the law is like in the UK, but I can think of no law in the United States where Alex Chandler is located that would grant either a dealer or the owner of an instrument any kind of proprietary interest in historical sales data obtained from publicly-available sources.
  5. This is an important point, though I think the pictures (so far not shared) would be the biggest problem for Alex. While there are contexts in which it's acceptable to scrape the web and reproduce images online (see Google Images, for example), grabbing a bunch of images off a website like Tarisio and redistributing them as part of an instrument database would certainly be in violation of copyright law. Sales price data is trickier -- information about individual sales is almost certainly unprotected, but collections of such data could still be copyrighted under particular circumstances. So, yeah, best to exercise caution.
  6. If you're lucky you may be able to find a copy of Paganini's Violin: its history, sound, and photographs, which includes a copy of The Strad's Il Canone poster. Looks like it's pretty hard to find, though. This website (which I've never used) shows it as "usually available for despatch within 2-3 working days," but "usually" could easily mean it's not in stock and they'll let you know after you've placed the order. Edit: here's another source, and for less money.
  7. Dissolve some shellac flakes into the hand sanitizer and call it French polish. *backs slowly out of room*
  8. And the user of Alex's program would likewise have to provide information about the target instrument's maker*, condition**, size, provenance, and other such relevant factors that can't necessarily be determined from pictures alone (if at all). -- * An AI might be able to guess the maker based on pictures in at least some cases, but even experts looking at instruments in person can disagree. How many instruments sold at auction are only "attributed to," "ascribed to," "probably by," or "possibly by" a given maker? Then there's that Del Gesu that sold at auction that nobody noticed was special except Stefan Hersh. ** The user would need to describe condition in a way that is consistent with how dealers determine condition.
  9. I haven't looked at your code, but I don't think that's the problem. I think your assumptions about the relationship between form and sales price are incorrect, so it really doesn't matter how good your code is you're still not going to get the results you are hoping for.
  10. I'm sorry if I came off too strong, but I think you'd have better luck training a neural network to identify violin models and/or "schools" than to determine value. You could feed your network a picture of an instrument and have it identify the instrument as, say, a "Strad" or "Guarneri" model with a high degree of confidence (for instruments that fit into such categories), but determining value is something else entirely. I'm suggesting the relationship between an instrument's appearance and its value is too tenuous to yield any useful predictions regardless of whatever spurious correlations might be found in your training set. I could be wrong.
  11. Not to be rude, but this seems about as likely to work as training a neural network to predict criminal behavior based on the shape of people's skulls. You're looking to derive a meaningful model from inputs that are not predictive of the outputs beyond whatever accidental relationships may be found in your training data.
  12. It's not necessarily behind a paywall. I sometimes get a login screen while browsing Tarisio, even on pages that aren't part of the paid Cozio Archive. I don't know exactly what triggers the login screen (I don't always get one), and logging in with a free account is enough to continue browsing, but it's pretty annoying and I have no idea why the folks at Tarisio think it's a good idea to do this.
  13. For the curious, here is an archive of Fritz Reuter's website.
  14. Some wood stats from Schwarz's first workbench book, Workbenches: From Design & Theory to Construction & Use. * The Janka scale indicates hardness.
  15. Nice. I've been reading Christopher Schwarz's other books on workbench design and they're both pretty good. I expect this one covers much of the same ground, but it looks nice enough that I may end up buying it anyway in spite of the free download (I prefer printed books anyway).