datasets:phoenix-annotations

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

datasets:phoenix-annotations [2018/09/03 19:35] (current)
Line 1: Line 1:
 +<​html>​
 +<h2 class="​sectionedit1"><​a name="​face_shape_annotations_for_the_rwth-phoenix_corpus"​ id="​face_shape_annotations_for_the_rwth-phoenix_corpus">​Face shape annotations for the RWTH-PHOENIX corpus</​a></​h2>​
 +<div class="​level2">​
  
 +<p>
 +The following archive available for <a href="/​public/​RWTH-PHOENIX-face-shape-annotations.tar.gz">​download</​a>​ contains face shape annotations – along with the corresponding images – for the seven signers who appear in the RWTH-PHOENIX corpus.
 +The archive contains one directory per signer, each of which contains an “images” subdirectory and a “points” subdirectory. All image files are 300&​times;​260 <acronym title="​Portable Network Graphics">​PNG</​acronym>​ images.
 +All face shape annotation files are <acronym title="​American Standard Code for Information Interchange">​ASCII</​acronym>​ files with the filename extension ”.pts”.
 +</p>
 +
 +<p>
 +<a href="/​public/​RWTH-PHOENIX-face-shape-annotations.tar.gz">​Download RWTH-PHOENIX face shape annotations here!</​a>​
 +</p>
 +
 +</​div>​
 +
 +<​h4><​a name="​distribution_of_the_annotated_images_among_the_signers"​ id="​distribution_of_the_annotated_images_among_the_signers">​Distribution of the annotated images among the signers</​a></​h4>​
 +<div class="​level4">​
 +
 +<p>
 +In total 369 images from the RWTH-PHOENIX corpus have been manually annotated with the face shape.
 +</p>
 +<div class="​table sectionedit2"><​table class="​inline">​
 +<tr class="​row0">​
 +<th class="​col0 leftalign">​ Signer ​     </​th><​th class="​col1">​ # annotated images </th>
 +
 +</tr>
 +<tr class="​row1">​
 +<td class="​col0">​ Signer01 </​td><​td class="​col1 centeralign"> ​ 61  </td>
 +</tr>
 +<tr class="​row2">​
 +<td class="​col0">​ Signer02 </​td><​td class="​col1 centeralign"> ​ 65  </td>
 +
 +</tr>
 +<tr class="​row3">​
 +<td class="​col0">​ Signer03 </​td><​td class="​col1 centeralign"> ​ 62  </td>
 +</tr>
 +<tr class="​row4">​
 +<td class="​col0">​ Signer04 </​td><​td class="​col1 centeralign"> ​ 51  </td>
 +
 +</tr>
 +<tr class="​row5">​
 +<td class="​col0">​ Signer05 </​td><​td class="​col1 centeralign"> ​ 35  </td>
 +</tr>
 +<tr class="​row6">​
 +<td class="​col0">​ Signer06 </​td><​td class="​col1 centeralign"> ​ 42  </td>
 +
 +</tr>
 +<tr class="​row7">​
 +<td class="​col0">​ Signer07 </​td><​td class="​col1 centeralign"> ​ 53  </td>
 +</tr>
 +</​table></​div>​
 +
 +</​div>​
 +
 +<​h4><​a name="​structure_of_a_face_shape_annotation_file_pts_file"​ id="​structure_of_a_face_shape_annotation_file_pts_file">​Structure of a face shape annotation file (&​quot;​.pts&​quot;​ file)</​a></​h4>​
 +
 +<div class="​level4">​
 +
 +<p>
 +For each image the face shape annotation consists of the ordered list of the positions of 38 facial landmarks. Each facial landmark
 +– or face shape point – has two coordinates in the coordinate system originating at the top-left corner of the image, with the x-axis pointing right
 +and the y-axis pointing down. The actual structure of a ”.pts” file is as follows:
 +</p>
 +<pre class="​code">​version:​ 1
 +n_points: 38
 +{
 +  x_1  y_1
 +  x_2  y_2
 +  ...
 +  x_38 y_38
 +}</​pre>​
 +
 +<p>
 +<img src="/​_media/​downloads/​phoenix-example.png"​ class="​medialeft"​ align="​left"​ title="​Face shape annotation example"​ alt="​Face shape annotation example"​ />
 +</p>
 +
 +<p>
 +A possible interpretation of the shape points as face parts is as follows (using zero-based indexing in the list of points of a ”.pts” file):
 +</p>
 +<div class="​table sectionedit3"><​table class="​inline">​
 +<tr class="​row0">​
 +
 +<th class="​col0 leftalign">​ Shape point indices ​ </​th><​th class="​col1">​ Face part </th>
 +</tr>
 +<tr class="​row1">​
 +<td class="​col0">​0-1-2 </​td><​td class="​col1">​ upper part of the left eyebrow</​td>​
 +</tr>
 +
 +<tr class="​row2">​
 +<td class="​col0">​3-4-5 </​td><​td class="​col1">​ upper part of the right eyebrow</​td>​
 +</tr>
 +<tr class="​row3">​
 +<td class="​col0">​6-7-8-9 </​td><​td class="​col1">​ eyelids of the left eye</​td>​
 +</tr>
 +
 +<tr class="​row4">​
 +<td class="​col0">​10-11-12-13 </​td><​td class="​col1">​ eyelids of the right eye</​td>​
 +</tr>
 +<tr class="​row5">​
 +<td class="​col0">​14-15 </​td><​td class="​col1">​ nasal ridge</​td>​
 +</tr>
 +
 +<tr class="​row6">​
 +<td class="​col0">​16-17 </​td><​td class="​col1">​ nasal base</​td>​
 +</tr>
 +<tr class="​row7">​
 +<td class="​col0">​18-19-20-21-25-24 </​td><​td class="​col1">​ upper lip</​td>​
 +</tr>
 +
 +<tr class="​row8">​
 +<td class="​col0">​18-23-22-21-26-27 </​td><​td class="​col1">​ lower lip</​td>​
 +</tr>
 +<tr class="​row9">​
 +<td class="​col0">​28-29-30-31-32-33-34-35-36-37 </​td><​td class="​col1">​ cheeks and chin</​td>​
 +</tr>
 +
 +</​table></​div>​
 +<p>
 +<br>
 +Note: The original images from the RWTH-PHOENIX corpus do not have square pixels and come in 210&​times;​260 image resolution.
 +Therefore the annotated data provided here should be scaled appropriately if the goal is to use them to build models that are relevant to
 +the entire corpus. Scaling both our images and our shape annotations by a factor 210/300
 +<​em>​in the horizontal direction only</​em>​ (x-coordinates) will do the trick!
 +</p>
 +
 +<​h4><​a name="​license"​ id="​license">​License</​a></​h4>​
 +<div class="​level4">​
 +
 +<p>
 +
 +<p>
 +<a rel="​license"​ href="​http://​creativecommons.org/​licenses/​by-nc-sa/​3.0/"><​img align="​left"​ alt="​Creative Commons License"​ style="​border-width:​0"​ src="/​_media/​downloads/​by-nc-sa.png"​ /></​a><​span xmlns:​dct="​http://​purl.org/​dc/​terms/"​ href="​http://​purl.org/​dc/​dcmitype/​Dataset"​ property="​dct:​title"​ rel="​dct:​type">&​nbsp Face shape annotations for the RWTH-PHOENIX corpus</​span>​ by <span xmlns:​cc="​http://​creativecommons.org/​ns#"​ property="​cc:​attributionName">​Thomas Hoyoux, <a href="​https://​iis.uibk.ac.at/">​IIS Lab</​a>,​ University of Innsbruck,</​span>​ are licensed under a <a rel="​license"​ href="​http://​creativecommons.org/​licenses/​by-nc-sa/​3.0/">​Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported License</​a>​. To view a copy of this license, visit <a href="​http://​creativecommons.org/​licenses/​by-nc-sa/​3.0/">​http://​creativecommons.org/​licenses/​by-nc-sa/​3.0/</​a>​ or send a letter to Creative Commons, 444 Castro Street, Suite 900, Mountain View, California 94140, USA.
 +</p>
 +
 +</p>
 +
 +</​div>​
 +
 +
 +</​div>​
 +
 +</​html>​
 +
 +
 +=== Acknowledgment ===
 +
 +This work was partially funded by the European Community'​s Seventh Framework Programme FP7/​2007-2013 (Specific Programme Cooperation,​ Theme 3, Information and Communication Technologies) under grant agreement no. 231424, [[http://​www.signspeak.eu/​|SignSpeak]].
datasets/phoenix-annotations.txt · Last modified: 2018/09/03 19:35 (external edit)