登录注册
Quick Links : Mindat手册The Rock H. Currier Digital LibraryMindat Newsletter [Free Download]
主页关于 MindatMindat手册Mindat的历史版权Who We Are联系我们于 Mindat.org刊登广告
捐赠给 MindatCorporate Sponsorship赞助板页已赞助的板页在 Mindat刊登 广告的广告商于 Mindat.org刊登广告
Learning CenterWhat is a mineral?The most common minerals on earthInformation for EducatorsMindat ArticlesThe ElementsThe Rock H. Currier Digital LibraryGeologic Time
搜索矿物的性质搜索矿物的化学Advanced Locality Search随意显示任何一 种矿物Random Locality使用minID搜索邻近产地Search Articles搜索词汇表更多搜索选项
搜索:
矿物名称:
地区产地名称:
关键字:
 
Mindat手册添加新照片Rate Photos产区编辑报告Coordinate Completion Report添加词汇表项目
Mining Companies统计会员列表Mineral MuseumsClubs & Organizations矿物展及活动The Mindat目录表设备设置The Mineral Quiz
照片搜索Photo GalleriesSearch by Color今天最新的照片昨天最新的照片用户照片相集过去每日精选照片相集Photography

Improving Mindat.orgcan we improve how "Rating" is defined for images?

24th Jul 2023 05:29 UTCDon Windeler

For a while now I’ve had an itchy question about the “Rating” aspect of photos on MinDat.  As ever, I appreciate the community-based aspect of this site and am not criticizing – much – but am more looking to improve.  Note comments here are not aimed at any user’s photos, rather the general process.

What determines what a photo receives as its “Rating” that is used for sorting images at a locality or other searches?  As a regular user, I do not see any way of influencing this unless it is by clicking on the image and adding to its view count – there is no option to “upvote” or “like” a relevant image.


Part of why I ask this is frustration with a locality I know pretty well, the Yerington district in western Nevada (https://www.mindat.org/loc-6712.html).  While there are many folks more expert than me, I did do my master’s thesis in the skarns near Ludwig from 1989-1992 and know a few things.  I’ve periodically sponsored the area.

If I visit the region and look at mineral photos and use the default view by Rating, the piece that comes up is a travertine attributed to the Casting Copper mine (MinID 83E-ERC).  Pretty sure this specimen was tagged here due to an earlier photo I posted of a similar small, crappy piece (CTH-4TL) that I collected and posted a year earlier.  (Note I just fixed the locality on my piece, so they won't show the same as of this post.)

My question is how this other piece became the top-rated image for the Yerington district?  The Yerington district was a major porphyry copper producer that was operating for close to a century and still has lots of reserves.  There’s a huge range of primary and secondary minerals for it, including some fascinating, well-studied skarns on the periphery.  The Ludwig area in particular is famous among knowledgeable regional collectors for its chartreuse vesuvianite crystals that reached 10+ cm in size.

Somehow, though, the top rated photo for the district is a secondary calcite that has basically nothing to do with the region, was one limited seam (as far as I’ve seen), and isn’t even that great of a photo.  If the “Rating” criteria can only be influenced by clicks, however, it’ll keep getting bumped farther up the list because it’s the top-rated image when someone clicks through to the region.  For Yerington, there are so many that would be better choices – personally I like PLF-TE5, because even though Rock left off the size you can see how big the vesuvianites got.


Please note I’m not campaigning for any of my photos here, most of which stink.  There are dozens of photos that better represent the area and deserve to show up as the top rated for the locality.  Looping back to my original question, how can the community influence image Rating?  If it’s only via clicking on a photo, the current system will reinforce selection of the images with the most clicks, even if they are irrelevant to a locality.

Upvoting... just another of those nice-to-have features for which we all should contribute to support somehow!

Cheers,

D.

 

24th Jul 2023 05:53 UTCJohn Christian

In the link to the Yerington District photos, where are their “ratings” visible?

24th Jul 2023 05:59 UTCDon Windeler

Ratings are never visible, which is one of the issues.  You just show the images for a location, mineral, or user and "Rating" is one of the criteria by which you can sort -- it's just not clear what value is used.

D.

24th Jul 2023 07:13 UTCDave Griffiths

There is a photo rating section where you get a random mineral type in a random locality to judge. The odd thing is that I only accidentally found this a few days ago, and now I can't find it again :) Maybe it needs to be more prominent.

24th Jul 2023 07:15 UTCDave Griffiths


Perhaps this is what is being used? It is unclear.

Edit: It's actually very fun and quite educational, to say the least - I think it should indeed be more prominent, which would increase the quality of photos seen overall.

24th Jul 2023 08:46 UTCDavid Von Bargen Manager

There are three criteria used to rate photos. A manager (and photo approval users) rating, the user rating (from revpicssel), and the view count. Ties in any level are decided by the next lower criteria.
Managers also have a manual override for photos shown on locality and mineral pages. Used for 800 location and 500 mineral pages.

There are 42K photos that have a manager rating, 50K user ratings.

The revpicssel is set up so you do not see any of your photos to rate and increase user rating by 1.

Upvoting is not used so you don't have cases where a photographer can "game" the system or work with another user to upvote a particular photo. (it's a bit much to save who upvotes various photos).

24th Jul 2023 10:27 UTCJolyon Ralph Founder

Rating algorithms here will have to change over time because what is an appropriate algorithm for 50 photos doesn't necessarily scale up to 50,000.

If you have a large number of photos that have been given a lot of rating scores over many years, that potentially can make it very hard for high quality new photos to be seen. 

So, things that can go into the rating system include:

Image resolution (higher resolution images preferred over low-res)

A rating for the photographer based on their contributions to the site. For example, how many POTDs they have had, their level of access, and overall number of complaints about images in the past.

How new the image is

How many 'flags' have been raised in approval (each of which should significantly decrease the score)

Votes from members are useful but need to be balanced with the age of an image - newer images won't have had the opportunity to receive as many votes.



We want new images to be given more of a chance than they are now.

25th Jul 2023 07:33 UTCHerwig Pelckmans

Jolyon Ralph Founder  ✉️

We want new images to be given more of a chance than they are now.

 If mindat really wants to do that, then I suggest that the default of the mineral gallery pages is changed:

Right now, the default "Order by" of all those pages is the "Rating". 
=>  this ordering does not make much sense, since most photos are not rated at all.

I suggest changing the default to "Newest first".
=> this makes a lot of sense, since older photos in general are of a lesser quality.

Hope you can make this happen sooner than later! :-)
Cheers, Herwig

25th Jul 2023 08:55 UTCJolyon Ralph Founder

Newest first isn't a great idea either.  

25th Jul 2023 10:12 UTCHerwig Pelckmans

Jolyon Ralph Founder  ✉️

Newest first isn't a great idea either.  

Please elaborate why you think it is not a great idea

Cheers, Herwig

25th Jul 2023 10:27 UTCDavid Von Bargen Manager

It's no better than ordering them randomly. Recent photos are not necessarily better than older ones.  

25th Jul 2023 10:43 UTCHerwig Pelckmans

David Von Bargen Manager  ✉️

Recent photos are not necessarily better than older ones.  

Depends on the kind of photos. In general, photos with a fov of (say) 5 mm or less have continuously (and dramatically) improved in quality in the last 20 years, mainly due to more and more people using stacking & stacking software. Stacking software has evolved too.

Considering that a lot of photos are being taken again, by the same photographer and of the same specimen, this also indicates that newer equipment and newer software produce better photos.

And considering that a lot of photos on mindat are photos made using stacking, I would say that, in general, more recent photos are better than older ones. Hence my plea for changing the default.

Cheers, Herwig

25th Jul 2023 11:59 UTCDave Griffiths

I agree that often ordering by newness can, but not always, result in good photos - but what might be more important is that it gives an idea of the vibrancy and activity of the site, which is also important I think.

25th Jul 2023 12:51 UTCDavid Von Bargen Manager

if that is important to you, you can filter galleries that way.

25th Jul 2023 12:59 UTCHerwig Pelckmans

David and Jolyon,
so far I have not seen an argument why the default should not be changed from "rating" to "newest first".
Dave just added another argument in favor of "Newest first".
If there are any arguments in favor of "Rating", please let us know.
Cheers, Herwig

25th Jul 2023 13:11 UTCDavid Von Bargen Manager

No, I didn't. If you like them to be in that order you can view them that way. That is not necessarily the best  way to view them or get the best photos first.
Newest first is just another sort of a random ordering of photos. 

Ordering them by ratings at least has some expert input as to the order that they should be displayed.

25th Jul 2023 14:50 UTCHerwig Pelckmans

David Von Bargen Manager  ✉️

Ordering them by ratings at least has some expert input as to the order that they should be displayed.

Of course I understand your point of view, David, and I agree that ordering them by rating makes some kind of sense. How much sense, depends on the number of photos that have been rated as well as when most of them were rated.

So, can you tell us how many mineral photos there are, and how many of those have been rated? The homepage states 1,287,370 photos, but I assume not all of those are mineral photos.

Out of curiosity, I tried to do the same for the mineral fluorite. 
The fluorite page mentions no less than 30333 photos(!) of this mineral in mindat.
Next, I did a "Search photos of fluorite" and on the page of the "Fluorite - Photo gallery", I chose for "Show", the option "Top rated images".

To my surprise, only 1 single photo shows to be a top rated image:
Does this mean there are more than 1 images that are rated, but this is "the best of the best"? Or does this mean there is only 1 image of fluorite that is rated?

And BTW, considering the low quality of the image, IMHO this photo does not even deserve to remain in the Public Galleries today. It might have been a great photo when it was uploaded to mindat (Approval date: 2007-02-19), but it most certainly is not today.

Cheers, Herwig

25th Jul 2023 14:57 UTCAlfredo Petrov Manager

"...considering the low quality of the image, IMHO this photo does not even deserve to remain in the Public Galleries today. It might have been a great photo when it was uploaded to mindat (Approval date: 2007-02-19), but it most certainly is not today."

Herwig, this photo must remain in the public galleries because it is the only photo we have of fluorite from this locality. It's not only about photo quality competition, there are also educational reasons for keeping a lot of the photos on Mindat.

25th Jul 2023 15:00 UTCDavid Von Bargen Manager

Herwig Pelckmans Expert  ✉️

So, can you tell us how many mineral photos there are, and how many of those have been rated? The homepage states 1,287,370 photos, but I assume not all of those are mineral photos.
 I answered that in my first response in this thread.
Total number of specimen photos1,066,157
Total number of locality photos114,330
Total number of other photos39,070
Total photos1,219,557  

This is available in statistics in the more button

25th Jul 2023 15:17 UTCHerwig Pelckmans

David Von Bargen Manager  ✉️

There are 42K photos that have a manager rating, 50K user ratings.
Okay, so in short: over 1,066,000 specimen photos, and maximum 92,000 of those have a rating at the moment. That is a bit less than 9%. 

It's probably more difficult to figure out when most of these ratings were given. Take for example the fluorite photo again. Since Alfredo pointed out it is the only photo of fluorite from that locality in mindat, I agree it should remain in the Public Galleries. But I disagree this photo deserves to be "top rated".

I fear that most rated photos have numbers lower than 500,000. If that is the case, the ratings are less meaningful than what you would like them to be.

Cheers, Herwig 

26th Jul 2023 16:35 UTCJolyon Ralph Founder

In general it's true that newer photos tend to be better than older photos, so let's choose five minerals at random and see which is the better choice, newest or default.

Let's take (mix of rare & common)

quartz
fluorite
bayldonite
cubanite
vesuvianite


Quartz -  Default order is *MUCH* better than Newest

Fluorite - I'd say it's a draw. The photos on 'Newest' are better, BUT we have the problem with newest that if someone uploads a large number of photos from the same locality in one go, then the 'newest' page becomes flooded with one locality.

Bayldonite - Default order wins for me

Cubanite - Default order way better

Vesuvianite - Again I'd say closer to a draw, the photos on 'default' are slightly better, but I'll be fair and call this one a draw.


----

I wasn't expecting the results to be so conclusive, but to me this indicates that it would be a very bad idea to switch the default to 'newest'

25th Jul 2023 13:26 UTCAlfredo Petrov Manager

04274410017060065817744.png
Here is a screenshot of the list you get when you click on "Order by" in any photo gallery. As David pointed out, one can already choose to order by newest first, or ratings, and several other ways.

I wonder whether our programmers could figure out a way for one of these choices to become the default preference of a user, so that the user doesn't need to reselect it again for each gallery visited? (Perhaps that's already possible, and I'm just not aware of the feature.) Meanwhile, I'm satisfied with the system as it currently operates. 

25th Jul 2023 13:51 UTCYuri Spektor

As a web developer, I think this should not be hard to implement, this could be done using cookies or the browser local storage, the last user's preference could be stored there and set as default next time the page is loaded

26th Jul 2023 16:25 UTCJolyon Ralph Founder

Yes, you can indeed already set this order 

https://www.mindat.org/device_settings.php


25th Jul 2023 17:00 UTCEd Clopton 🌟 Expert

Seems like the main components of an automatic rating formula to be applied to all specimen photos, summarizing from Jolyon above, are:

Popularity over time:  a measure of how appealing, interesting, and/or useful users have found the photo, calculated more or less as (how often the photo has been viewed) divided by (how long it has been available) .  If desired, this could be scaled in various ways to favor newish vs. brand-new, new vs. old, etc.  POTDs get an automatic bump with this measure, since they generally get 200-400 views during just their 24 hours in the spotlight.

Expert evaluation:  (however that is done)  Since less than 10% of our photos have been given a rating, it can't be applied evenly across all photos and should not be weighted very heavily.  Some excellent shots that missed getting an expert rating would be at a disadvantage to middling ones that happened to become rated.

Image resolution:  In theory a higher-resolution image should be of higher quality and deserving of a higher rating, but a weak photo taken at high resolution is still a weak photo, and I don't know how an automatic rating formula could correct for that.

Not a finished proposal, just food for thought.

25th Jul 2023 18:29 UTCAlfredo Petrov Manager

Perhaps an AI could be trained to rate photos, and then all million remaining unrated photos could be finished in a few days  ;))

26th Jul 2023 12:22 UTCBill Hamel 🌟

Some thoughts on the above approach on “improving photo ratings.”

One of the most important and valuable things about Mindat is that it credibly preserves historical and scientific information in an easily accessible format.  In the mineral community it is considered the “gold standard” of web based technical information on minerals and mineral localities.

Therefore, I would propose deleting the criteria “Popularity over time,” simply because these factors seem to favor a “social media type approach,” which could weaken Mindat’s technical credibility. 

I would double weight and add the following criteria:

Does the picture have a write-up? If it does not, it should receive a lower (or no) rating. Pictures without a write-up are a missed opportunity to explain & preserve important data.

Does the picture convey important and/or relevant scientific information?  There are a lot of pictures with write-ups in Mindat that convey unique information that isn’t available (or easily available) from other sources.  Many of these pictures do not have aesthetic subjects and may not get a lot of views yet are invaluable resources.

Does the picture convey important and/or relevant historical information? Similar to the above, there are a lot of pictures with write-ups in Mindat that convey unique historical data that isn’t available (or easily available) from other sources.  Many of these pictures do not have aesthetic subjects and may not get a lot of views yet are invaluable resources.

My two cents.

Thanks as always - Bill
 
and/or  
Mindat Discussions Facebook Logo Instagram Logo Discord Logo
版权所有© mindat.org1993年至2024年,除了规定的地方。 Mindat.org全赖于全球数千个以上成员和支持者们的参与。
隐私政策 - 条款和条款细则 - 联络我们 - Report a bug/vulnerability Current server date and time: 2024.5.14 16:27:01
Go to top of page