does anyone have experience with this software? I tried messing around and get errors...
strangebeta dot com
It was also written about in one of the mags this month or last.
strangebeta - anyone?
(12 posts) (4 voices)-
Posted 2 years ago #
-
gong
Posted 2 years ago # -
I also get a ton of errors.
I haven't read the article yet, but if you read the technical paper I bet this discussion might get a little interesting.
Commentary on section 5 would interest me.
Also, as far as I can tell, a route is defined in this system entirely by type and directionality of hand holds. The route is then either combined with another route or varied mathematically. The output is in the same format as the input.
It seems to me that if I were to give anyone in this forum an output that read: Right hand pinch, left hand sidepull, right hand gaston, etc... You could create quite a large number of routes that fit the description, who good they would be would still depend on you.
So are we talking about a software that will tell me that I can use different types of holds with different orientations? I'm having a hard time telling how this is different that random hold selection coupled with spinning for random orientation.
If not for the claims on the site I would readily assume that this is merely a mathematical exercise but now I am confused.
Posted 2 years ago # -
Hey guys, Steve from the Crux Climbing Gym let me know that some people were discussing StrangeBeta here. I'm actively working on StrangeBeta.com, fixing bugs and working on ways to improve the algorithms. I'm interested in any thoughts you have on the way the software works currently and how it might work better.
First off, if you're having errors let me know---either here or by email (caleb.phillips <shift>+2 colorado.edu) and I can probably fix them. Particularly, let me know what you were doing that caused the error so that I can reproduce it. It's "research grade" software, but it still should work better than producing a "ton of errors"!
--
Caleb PhillipsPosted 2 years ago # -
Jason, you're right - the way it works now is that it randomly varies the input routes. However, it's a /bit/ "smarter" than just random hold selection and randomly spinning holds. Mostly, it's that it uses a very special "kind" of random rearrangement that attempts to preserve the style of the input. It does this by rearranging sub-sequences instead of individual holds. This is still pretty dumb, which is why the output must be interpreted by an experienced setter to make a good problem.
Currently (today even), I'm working on adding some machine-learning algorithms to the site. This would have the effect of "learning" what a climb looks like, and probably what a crux looks like. Then the program can both smooth transitions between rearranged chunks, and generate its own interesting subsequences to insert. The upshot is that the output will be more sophisticated and take a stronger creative roll.
Posted 2 years ago # -
I bet it could write some great pop songs too
Posted 2 years ago # -
That's not a very high bar ;)
Posted 2 years ago # -
seriously though .. I have not enough time to get my own ideas on the wall.. although it does sound interesting I'll admit. I would certainly give it a try before I really gave it a gong.
Posted 2 years ago # -
Caleb,
Alright, random wasn't fair. But I wasn't trying to demean (or even discuss) the algorithm. My interest lies more in the assumptions.
You seem to be readily assuming that neither the line of the route nor the foothold usage is highly relevant when you output only the information that you do.
Presumably you are leaving these factors in the hands of the "experienced setter." I understand this choice in that your problem would become exponentially harder to define with these added factors but then I don't understand your success criteria. The program is considered successful if the routes generated are at least as good as routes set traditionally, right? So in tis scenario, how much of a roll is the program actually playing in this outcome? Basically the strongest claim you can make is that the program is able to generate outputs that a route setter is able to use, no?
Posted 2 years ago # -
Jason, all of your assertions about the experimental design are correct. I also can't claim that my results in the BRC allow me to extrapolate to any other gym or any other setter. What I found working with the setters at the BRC was that setters substantially enjoy the creative aspects of route-setting and are highly skeptical of anything that would lessen their creative control, yet, they did admit that having a plan to "go off of" made their task somewhat easier (and the result for the 5.11 climbs was better in the study we did). This comports to my goal of creating something to assist setters when they don't particularly feel like coming up with a route "from scratch". I imagine a setter who must put up three routes at 7am, with a raging hangover (for instance) appreciating such a utility.
The assumption that foot placements can follow from hands (i.e. they don't need to be specified in the software) is something that all setters I've talked to have agreed with - they say they typically set feet to accommodate hands. Is this your experience as well? And the line/steepness etc., yes, those are left "open", but maybe not in future versions. I'm still learning what the important criteria to specify are. My feeling was that keeping things as "free form" as possible in the beginning would give me room to learn what the right level of specification is. In this way, I've deviated from prior work in dance choreography where researchers fully specified the orientations of 23 main joints in a dancer's body. Which, while cool, introduces substantial barriers for adoption (imagine needing a special studio and a bunch of computers and technicians just to input a route into my software).
Posted 2 years ago # -
Alright, I'm catching on now. If you take the setting as art vs. craft situation and accept that there are situations where art doesn't fit (time constraints, lack of interest, etc.) then presumably any usable tool deserves appreciation. Since we set as volunteers out here this situation never arises and I tend to forget.
The foot placement question is actually tricky. As you've worded it it is certainly true. If you have usable hands on the wall any setter can place feet to accommodate. But it is certainly not the case that foot placement must follow in this way. Foot placement and line tend to be what differentiate a decent route from a great route. That's why I was curious about the success criteria. It just didn't seem to follow.
Anyway, it's certainly an interesting project and I understand (from the programming side) why you made the choices that you did. I was only disappointed at first because, based on the initial description, it sounded like it might be a really useful learning tool for new setters.
Posted 2 years ago # -
Interesting point about craft versus art, I hadn't thought of it like that.
Using it as a learning tool for new setters is an interesting application I've been thinking about. However, right now I can't rubber-stamp it for this purpose with much confidence. There also seems to be a confounding issue that we came across in our study that a route "transcribed" by a given person might be difficult to use by another person. This stems from what I observe to be personal idioms in the way people describe a route's moves. For instance, "out to 90 degree jug", might have a very specific meaning to the person who wrote it, and might be ambiguous to everyone else.
I'm hoping that the addition of learning algorithms will make the tool more useful overall. It might also lead to some interesting observations about the aesthetic of setting itself (i.e., what makes a route "good"). Ultimately this is just a side project for me---an interesting diversion from my main work (and an excuse to go climbing and say I'm "working" ;)). So, to the extent that it's useful at all, I'm stoked.
Posted 2 years ago #
Reply
You must log in to post.