Skip to content

Commit

Permalink
Updated FAQ and started dev information
Browse files Browse the repository at this point in the history
  • Loading branch information
jonasob committed Nov 18, 2014
1 parent 00bebca commit 80cc65a
Show file tree
Hide file tree
Showing 6 changed files with 153 additions and 20 deletions.
10 changes: 7 additions & 3 deletions _includes/footer.html
Original file line number Diff line number Diff line change
Expand Up @@ -5,15 +5,19 @@
<div class="col-md-8">
<ul class="list-inline">
<li>
<a href="#home">Home</a>
<a href="/">Home</a>
</li>
<li class="footer-menu-divider">&sdot;</li>
<li>
<a href="#about">About</a>
<a href="/faq/">FAQ</a>
</li>
<li class="footer-menu-divider">&sdot;</li>
<li>
<a href="#services">Services</a>
<a href="/privacy/">Privacy Policy</a>
</li>
<li class="footer-menu-divider">&sdot;</li>
<li>
<a href="/dev/">Developer</a>
</li>
<li class="footer-menu-divider">&sdot;</li>
<li>
Expand Down
4 changes: 2 additions & 2 deletions _includes/navigation.html
Original file line number Diff line number Diff line change
Expand Up @@ -18,10 +18,10 @@
<a href="#features">About</a>
</li>
<li>
<a href="#">FAQ</a>
<a href="/faq/">FAQ</a>
</li>
<li>
<a href="#">Developer</a>
<a href="/dev/">Developer</a>
</li>
<li>
<a href="#contact">Contact</a>
Expand Down
1 change: 0 additions & 1 deletion _layouts/faq.html
Original file line number Diff line number Diff line change
Expand Up @@ -10,7 +10,6 @@
<p><a href="#general">General questions</a></p>
<p><a href="#matching">Image matching</a></p>
<p><a href="#database">Database information</a></p>
<p><a href="#privacy">Privacy policy</a></p>
<p><a href="#contact">Contact information</a></p>
</div>
<div class="col-md-2">
Expand Down
9 changes: 9 additions & 0 deletions dev.html
Original file line number Diff line number Diff line change
@@ -0,0 +1,9 @@
---
layout: plain
title: Elog.io - Developer information
root: "../"
---

<div class="faq">

</div>
93 changes: 79 additions & 14 deletions faq.html
Original file line number Diff line number Diff line change
Expand Up @@ -19,7 +19,7 @@
<p>
We currently have information about photographs from
<a href="http://commons.wikipedia.org">Wikimedia Commons</a>, a collection
of approximately 23 million images (November 2014). The photographs on
of approximately 23 million photographs (November 2014). The photographs on
Wikimedia Commons in turn come from a wide variety of sources: the authors
themselves, imported from <a href="http://flickr.com">Flickr</a>, or other
sources.
Expand All @@ -33,17 +33,28 @@
<p><strong>So only openly licensed photos then?</strong></p>
<p>
Yes. We strongly believe in the open ecosystem and believe that answering
the basic question: is this openly licensed? for any image is important to
the basic question: is this openly licensed? for any photograph is important to
facilitate the ecosystem. For photographs that you can identify through
Elog.io, you can be reasonably certain that they are indeed licensed in the
way indicated (public domain, a Creative Commons license or similar).
</p>

<p><strong>Can I include my own photos?</strong></p>
<p>
We'd love to include your photos, or any collections that you're responsible
for, so we can make them findable through Elog.io. Please
<a href="#contact">contact us</a> for more information. Please note that at
this time, we're primarily looking at including larger collections of openly
licensed photographs, and the more work you're willing to do yourself on
making them available to us - or creating a project together to do the work -
the more likely it is that we can include them quickly.
</p>

<p><strong>Why is this even important?</strong></p>
<p>
From a practical perspective, knowing who authored an image and what terms
From a practical perspective, knowing who authored a photograph and what terms
and conditions it's available under is important in order to understand how
you may use an image, and what attribution you must provide when you do.
you may use an photograph, and what attribution you must provide when you do.
</p>
<p>
What we believe is that having this information available also place the
Expand All @@ -56,12 +67,20 @@
sense take precedence.
</p>

<p><strong>What's your privacy policy?</strong></p>
<p>
We take every precaution to keep your information safe, and we don't
purposefully collect any more information than we need to do the actual
processing on your behalf. We also remove information as soon as we're
done with it. You can read more in our <a href="/privacy/">Privacy Policy</a>.
</p>

<a name="matching"></a><h2>Image matching</h2>
<p><strong>Why doesn't Elog.io match X?</strong></p>
<p>
The matching algorithm that powers Elog.io is based on
<a href="http://blockhash.io">Blockhash</a>, a free and open source software
implementation of an image matching algorithm. In our research, we've found
implementation of an photograph matching algorithm. In our research, we've found
that this algorithm gives the best possible matches and it regularly
outperforms many other algorithms. That doesn't mean it's perfect though!
</p>
Expand All @@ -79,8 +98,8 @@
The Blockhash algorithm also doesn't deal well with modifications to
photographs. We've set the bar at verbatim re-use. Photographs that are
resized or are changed from JPG to PNG or similarly, but still retain the
same aspect ratio, and the same image, are the photographs that the algorithm
is designed to match. The moment you crop the image, add borders, change the
same aspect ratio, and the same photograph, are the photographs that the algorithm
is designed to match. The moment you crop the photograph, add borders, change the
colors, or do other manipulations, you're creating a derivative work, and
Elog.io will most likely not match this.
</p>
Expand All @@ -95,17 +114,17 @@
<ul>
<li>They should not be encumbered by patents, and we should be able to implement them in Free and Open Source Software, like Elog.io.</li>
<li>They should work well in a browser, meaning that they should be reasonably fast and easy to implement in JavaScript.</li>
<li>They should generate a hash value, with the particular characteristics that small changes to the image (resize, etc) should result in small changes to the hash.</li>
<li>Over a random sample of images, they should generate no or very few false positives (images that are matched against each other despite being different) and no or few false negatives (images that doesn't match, despite that they are the same).</li>
<li>They should generate a hash value, with the particular characteristics that small changes to the photograph (resize, etc) should result in small changes to the hash.</li>
<li>Over a random sample of photographs, they should generate no or very few false positives (photographs that are matched against each other despite being different) and no or few false negatives (photographs that doesn't match, despite that they are the same).</li>
<li>In general, they should prioritise generating no false positives, since a false positive is worse than a false negative.</li>
</ul>
<p>
When we evaluate an algorithm, we generate hashes against a
representative set of images that we have on file. We do two tests. In the
first test we transform the images (resize, format change, cropping,
add borders, rotate), and we test to see how much we can transform an image
representative set of photographs that we have on file. We do two tests. In the
first test we transform the photographs (resize, format change, cropping,
add borders, rotate), and we test to see how much we can transform a photograph
while still matching the original. In the second test we generate hashes
for a large amount of images and we cross compare them against each other.
for a large amount of photographs and we cross compare them against each other.
We calculate the hamming distance (number of bits that differ) between the
hashes and find out how often we would get false positives.
</p>
Expand All @@ -117,9 +136,55 @@
The easiest way is that you fork <a href="http://github.com/commonsmachinery/blockhash">Blockhash on Github</a> and create an implementation with your
algorithm. You can then submit a pull request and/or create an issue with
a reference to your fork, to allow us to pull your code to test it.
</p>

<a name="database"></a><h2>Database information</h2>
<a name="privacy"></a><h2>Privacy policy</h2>
<p><strong>How many photographs are in your database?</strong></p>
<p>
As of the 18th of November 2014, we have information of 22,412,293 photographs in the
Elog.io catalog. These photographs are exclusively from Wikimedia Commons
and as such represent a very wide variety of photographs, most of which are
already used elsewhere in Wikipedia.
</p>

<p><strong>What information do you store?</strong></p>
<p>
We store information about the author, title and license information about
each photograph. In addition to this, we record the URLs of the photograph
as well as at least one thumbnail or smaller resolution photograph. This is the
information that we get from Wikimedia Commons.
</p>
<p>
We supplement the information from Wikimedia Commons with our own
calculation of a <a href="http://blockhash.io">Blockhash</a> value for each
photograph.
</p>


<a name="contact"></a><h2>Contact information</h2>
<strong><p>Who runs this?</p></strong>
<p>
The initial work of Elog.io was provided by <a href="http://commonsmachinery.se">Commons Machinery</a>, funded by the <a href="http://shuttleworthfoundation.org">Shuttleworth Foundation</a>. The project is run as a free and open source
project by <a href="http://about.me/jonaso">Jonas Öberg</a>. Hosting is
currently provided Commons Machinery.
</p>

<strong><p>Can I email you?</p></strong>
<p>
We'd love to hear from you! You can email us at hello@elog.io.
</p>

<strong><p>What other ways can I get in touch?</p></strong>
<p>
Email is really the key to our hearts, but you can also reach out to us
on <a href="http://twitter.com/elog_io">twitter (@elog_io)</a> or by
sending post to us at:
</p>
<p>
Elog.io<br />
Åsgatan 5<br />
646 32 Gnesta<br />
SWEDEN
</p>

</div>
56 changes: 56 additions & 0 deletions privacy.html
Original file line number Diff line number Diff line change
@@ -0,0 +1,56 @@
---
layout: plain
title: Elog.io - Privacy Policy
root: "../"
---

<div class="faq">
<h2>Temporary information</h2>
<p>
When Elog.io is turned on, it searches the web pages you're viewing for
images and queries the Elog.io servers for information about those images.
The information it sends to the Elog.io servers consists of the image URLs
and (if you click "Query Elog.io", a calculated hash value representing
the image). Your connection to Elog.io is sent through a load balancer that
removes the originating IP address. This means that from the Elog.io server's
side, it will only see a sequence of URLs and hash values, without knowing
the relationship between them or the origin of each.
</p>
<p>
The Elog.io extensions don't send any other identifying information to
the servers, and no information is stored in the Elog.io database about
your request after it has been processed.
</p>
<p>
Elog.io is explicitly deactivated when you're browsing in a private or
incognito sesssion.
</p>

<h2>Stored information</h2>
<p>
If you submit feedback through the Elog.io extension, or report information,
we will store the information you provided for our own handling. We will
store this information for as long as we need it to act on your report or
feedback.
</p>
<p>
Our servers are hosted in London, United Kingdom.
</p>

<h2>Catalog information</h2>
<p>
The Elog.io catalog contain information about openly licensed photographs
published through Wikimedia Commons. If you have included or contributed to
photographs on Wikimedia Commons, your information is also included in
Elog.io.
</p>

<h2>Contact</h2>
<p>
Should you have any questions or concerns in relation to our handling
of your information, or the information of any 3rd party, we encourage
you to reach out to us through any of the <a href="/faq/#contact">contact
methods</a> we have available.


</div>

0 comments on commit 80cc65a

Please sign in to comment.