User talk:MfortyoneA

Welcome to Wikimedia Commons, MfortyoneA!

First steps tutorial Our first steps tour and our frequently asked questions will help you a lot after registration. They explain how to customize the interface (for example the language), how to upload files and our basic licensing policy (Wikimedia Commons only accepts free content). You don't need technical skills in order to contribute here. Be bold when contributing and assume good faith when interacting with others. This is a wiki. Getting help More information is available at the community portal. You may ask questions at the help desk, village pump or on IRC channel #wikimedia-commons (webchat). You can also contact an administrator on their talk page. If you have a specific copyright question, ask at the copyright village pump.

Goodies, tips, and tricks Put Babel boxes on your user page so others know what languages you can speak and indicate your graphic abilities. You can find the files you have uploaded in your Uploads. Please sign your name on Talk pages by typing ~~~~. To link to an image page without embedding the image, type: [[:File:Foo.jpg]], which produces: File:Foo.jpg. You may enable gadgets (custom features) for your account. There are several tools to upload files. Have you edited Wikipedia before? This guide might be useful. Made a mistake? Do you want to have a file renamed or moved? Use the move link and wait for an authorized user to rename it. Do you want to have your recently uploaded picture removed? Tag it as {{speedy|reason here}}. For more information read the deletion guidelines.

-- Wikimedia Commons Welcome (talk) 05:16, 10 March 2017 (UTC)

Annotations

You seem to be adding annotations of some pretty trivial things to photos whose main interest lies elsewhere. Is this part of some plan, or just messing around? - Jmabel ! talk 22:06, 24 March 2017 (UTC)

is this inappropriate? my thinking was links to and from wikipedia are a good thing, wiki projects's ability to function as a visual dictionary, resource for machine learning (e.g. being such a large, popular publicly available data-set). Is there a better way to scratch that itch. Is there a way to prioritize annotations - I notice there is a 'style' option MfortyoneA. It's also possible that in getting to know the content 'trivial' links could be later refined, e.g. 'tree' -> 'specific species of tree', arches->specific architectural jargon (if another user knows a bit more..) (talk) 03:12, 25 March 2017 (UTC)

It's not exactly inappropriate, but it can amount to stating the obvious. When someone is looking at an image, annotations draw their attention to things that might be of interest. These are probably harmless; they would possibly be useful if they were about things a bit more specific than, say, "lightpost". The place where they would definitely become a liability would be to add annotations like this to an image that has significant annotations, because the significant ones are likely to get lost in the shuffle by anyone viewing the image. - Jmabel ! talk 05:14, 25 March 2017 (UTC)

ok thanks for the feedback. I notice most images *dont* have annotations but perhaps i'll focus on looking for unusual bits of jargon. I'll dig around a bit more, I would hope the style could be used to de-emphasize. I notice the underlying platform supports arbitrary shapes too rather than just rectangles MfortyoneA (talk) 07:44, 25 March 2017 (UTC)

Just rectangles, which is kind of inconvenient. - Jmabel ! talk 16:05, 25 March 2017 (UTC)

At least your annotations should be correct, e.g. in File:2015-02-08 Molchowsee 206.jpg: this is a lake, so there is no "river bank", in File:Der Iterbach.jpg: no, this is a brook, no river, and no "ford" but just a few stones in the water, or in File:New Holland 110-90.jpg: no, this is not the engine, it's just the engine hood, etc.

But as Jmabel wrote above: its completely useless to mark all (or even: some) lake shores, pillars or road sides on in the millions of images on Commons. It would give sense to annotate things that are not obvious to be seen in the image, e.g. on the image of a lake the name of a village, a castle, or a lock near to it. But what in the image the lake shore is, what the reed belt and what the water surface, is completely clear to anybody without annotation - and a waste of time to add such annotations to the image. --Reinhard Kraasch (talk) 12:40, 14 April 2017 (UTC)

machine learning needs thousands ,or millions, of examples :) These things are obvious to humans but not obvious to machines. It would be nice if there was an official way to use wikimedia commons as a labelled image database. Labelling also makes the images available to later articles. I think increasing link density makes the resource more useful MfortyoneA (talk) 12:45, 14 April 2017 (UTC)

This is no aim for the annotations in Commons, they are to guide people and not to improve some hypothetical machine learning project. But even if this would be an aim, the annotations would have to be correct and not just mere guesses. --Reinhard Kraasch (talk) 12:51, 14 April 2017 (UTC)

sure, you are at liberty to correct any mistakes. I have replaced river bank with bank (geography). Thats the beauty of wiki, anyone can improve it. Even the mundane labelling can be refined (e.g. types of vehicle, more accurate names of architectural features). MfortyoneA (talk) 12:57, 14 April 2017 (UTC)

With a closer look on your work I estimate that it is in general rather misleading the image viewers than giving any useful information, I will revert your modifications and request you to stop it. --Reinhard Kraasch (talk) 13:00, 14 April 2017 (UTC)

Machine learning is not 'hypothetical'. ML is changing the world. We have a few companies with an advantage in proprietary data; they will dominate the world unless we have open datasets. Imagery from domestic and urban scenes will be used to guide robots in these environments (delivery, self-driving cars, domestic cleaning). Images of plants will be used to guide agricultural robots (planting, harvesting, pest-control). If the labelling looks un-natural now, please help convince the project maintainers to add some options for emphasis. The annotations have a style-parameter - this could easily be used to guide attention. I can't think of any better data-set than Wikipedia & by association commons - the fact that wikipedia doesn't just have 'names', it has detailed information explaining how things relate to eachother. MfortyoneA (talk) 13:02, 14 April 2017 (UTC)

Again: Within Commons this is not what annotations are thought for. If you want to improve machine learning you are free either to do this outside of Commons or to discuss the project aims of Commons at the appropiate pages. Right now these annotations are by all means unwanted, and if you continue to do so I will block your account. --Reinhard Kraasch (talk) 13:06, 14 April 2017 (UTC)

instead of deleting all these links (any idea how frustrating it is to have someone come along and delete everything?!!!!!), couldn't you help campaign for an appropriate style addition to the wikimedia commons software? It should be so easy to just change the emphasis. The images are already there, and for the sake of a few hundred additional bytes, their utility to humanity is enhanced. MfortyoneA (talk) 15:43, 14 April 2017 (UTC)

I know it's frustrating, and first hesitated to do so. But almost every annotation was either wrong or misleading, so I had no choice. It would have been better you had asked in advance (e.g. at the Village pump) instead of just started.

I guess you still do not understand what Commons and its user interface is meant for: It is for people, not for machines. Any information that is assisting people in using and understanding the images (e.g.: translating image descriptions, better categorize the images or geocode them) is appreciated, but anything that is misleading the users is not. When you want to use Commons images for training image interpretation algorithms you are free to do so, but not at the cost of degrading the usability of the images. BTW I do not understand what you mean with "appropriate style addition to the wikimedia commons software". Any modification of the software itself would have to be discussed with the developers and the Wikimedia Foundation. If you can convince them of whatever it is, it will be built in the software and hopefully be used by the Commons users. But we already have a lot of features and add-ons that are hardly used (e.g. the mentioned multilangual image descriptions, categories etc. etc.), the problem is the missing manpower for the current jobs, not the lack of ideas to spend additional time on. --Reinhard Kraasch (talk) 16:06, 14 April 2017 (UTC)

I suggest that if you want to proceed with this, you propose setting up a mechanism parallel to, but distinct from, ImageNote. You could then create a similar tool for your own purpose, without adding a bunch of annotations that direct the eye to minor details, and (this is the biggest issue) potentially prevent users from noticing significant ImageNotes. - Jmabel ! talk 17:58, 14 April 2017 (UTC)

ok I will look into the platform. MfortyoneA (talk) 19:25, 14 April 2017 (UTC)