Skip to content

Commit f23226c

Browse files
albaniealbanie
committedApr 7, 2017
added models page
1 parent 25f6a18 commit f23226c

File tree

5 files changed

+396
-1
lines changed

5 files changed

+396
-1
lines changed
 

‎css/custom.css

+9
Original file line numberDiff line numberDiff line change
@@ -16,3 +16,12 @@ footer {
1616
margin-top: 3% ;
1717
text-align: center ;
1818
}
19+
20+
.pretrained-models {
21+
font-size: 0.85em;
22+
}
23+
24+
.citation {
25+
color: orange;
26+
font-size: 0.9em;
27+
}

‎curriculum-vitals.html

+3
Original file line numberDiff line numberDiff line change
@@ -59,6 +59,9 @@ <h5 id="subheader">distinctly average computer vision researcher</h5>
5959
<li class="navbar-item">
6060
<a class="navbar-link" href="research.html">Research</a>
6161
</li>
62+
<li class="navbar-item">
63+
<a class="navbar-link" href="models.html">Models</a>
64+
</li>
6265
<li class="navbar-item">
6366
<a class="navbar-link" href="curriculum-vitals.html">CV</a>
6467
</li>

‎index.html

+3
Original file line numberDiff line numberDiff line change
@@ -58,6 +58,9 @@ <h5 id="subheader">distinctly average computer vision researcher</h5>
5858
<li class="navbar-item">
5959
<a class="navbar-link" href="research.html">Research</a>
6060
</li>
61+
<li class="navbar-item">
62+
<a class="navbar-link" href="models.html">Models</a>
63+
</li>
6164
<li class="navbar-item">
6265
<a class="navbar-link" href="curriculum-vitals.html">CV</a>
6366
</li>

‎models.html

+375
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,375 @@
1+
<!DOCTYPE html>
2+
<html lang="en">
3+
<head>
4+
<meta charset="utf-8">
5+
<title> S.M. Albanie</title>
6+
<meta name="description" content="">
7+
<meta name="author" content="">
8+
9+
<!-- Mobile Specific Metas
10+
–––––––––––––––––––––––––––––––––––––––––––––––––– -->
11+
<meta name="viewport" content="width=device-width, initial-scale=1">
12+
13+
<!-- FONT
14+
–––––––––––––––––––––––––––––––––––––––––––––––––– -->
15+
<link href="//fonts.googleapis.com/css?family=Raleway:400,300,600"
16+
rel="stylesheet" type="text/css">
17+
18+
<!-- CSS
19+
–––––––––––––––––––––––––––––––––––––––––––––––––– -->
20+
<link rel="stylesheet" href="css/normalize.css">
21+
<link rel="stylesheet" href="css/skeleton.css">
22+
<link rel="stylesheet" href="css/general.css">
23+
<link rel="stylesheet" href="css/custom.css">
24+
25+
<!-- Favicon
26+
–––––––––––––––––––––––––––––––––––––––––––––––––– -->
27+
<link rel="icon" type="image/png" href="images/favicon.png">
28+
29+
<script>
30+
(function(i,s,o,g,r,a,m){i['GoogleAnalyticsObject']=r;i[r]=i[r]||function(){
31+
(i[r].q=i[r].q||[]).push(arguments)},i[r].l=1*new Date();a=s.createElement(o),
32+
m=s.getElementsByTagName(o)[0];a.async=1;a.src=g;m.parentNode.insertBefore(a,m)
33+
})(window,document,'script','https://www.google-analytics.com/analytics.js','ga');
34+
35+
ga('create', 'UA-82913504-1', 'auto');
36+
ga('send', 'pageview');
37+
38+
</script>
39+
40+
</head>
41+
<body>
42+
43+
<!-- Primary Page Layout
44+
–––––––––––––––––––––––––––––––––––––––––––––––––– -->
45+
<div class="container">
46+
<div class="row">
47+
<div class="ten columns" style="margin-top: 25%">
48+
<h4>Samuel Albanie</h4>
49+
<h5 id="subheader">distinctly average computer vision researcher</h5>
50+
51+
<div class="navbar-spacer"></div>
52+
<nav class="navbar">
53+
<div class="container">
54+
<ul class="navbar-list">
55+
<li class="navbar-item">
56+
<a class="navbar-link" href="index.html">Me</a>
57+
</li>
58+
<li class="navbar-item">
59+
<a class="navbar-link" href="research.html">Research</a>
60+
</li>
61+
<li class="navbar-item">
62+
<a class="navbar-link" href="models.html">Models</a>
63+
</li>
64+
<li class="navbar-item">
65+
<a class="navbar-link" href="curriculum-vitals.html">CV</a>
66+
</li>
67+
<li class="navbar-item">
68+
<a class="navbar-link" href="https://github.com/albanie">Github</a>
69+
</li>
70+
</ul>
71+
</div>
72+
</nav>
73+
74+
<br>
75+
<h5 id="grimaces">Emotion Recogntion</h5>
76+
77+
The emotion recognition models listed below were trained to produce the results
78+
described in Table 2 and Table 3 of the paper,
79+
<a href="http://www.robots.ox.ac.uk/~albanie/publications/albanie16learning.pdf">
80+
Learning Grimaces By Watching TV</a>:
81+
</p>
82+
83+
<table class="pretrained-models">
84+
<thead>
85+
<tr>
86+
<th>model</th>
87+
<th>pretraining</th>
88+
<th>training</th>
89+
<th>model</th>
90+
<th>pretraining</th>
91+
<th>training</th>
92+
</tr>
93+
</thead>
94+
<tbody>
95+
<tr>
96+
<td><a href="http://www.robots.ox.ac.uk/~vgg/data/facevalue/alexnet-face-bn.mat">
97+
alexnet-face-fer</a></td>
98+
<td>vgg faces</td>
99+
<td>fer</td>
100+
<td><a href="http://www.robots.ox.ac.uk/~vgg/data/facevalue/alexnet-face-fer-bn.mat">
101+
alexnet-face-sfew</a></td>
102+
<td>vgg faces, fer</td>
103+
<td>sfew</td>
104+
</tr>
105+
<tr>
106+
<td><a href="http://www.robots.ox.ac.uk/~vgg/data/facevalue/vgg-m-face-bn.mat">
107+
vgg-m-face-fer</a></td>
108+
<td>vgg faces</td>
109+
<td>fer</td>
110+
<td><a href="http://www.robots.ox.ac.uk/~vgg/data/facevalue/vgg-m-face-bn.mat">
111+
vgg-m-face-sfew</a></td>
112+
<td>vgg face, fer</td>
113+
<td>sfew</td>
114+
</tr>
115+
<tr>
116+
<td><a href="http://www.robots.ox.ac.uk/~vgg/data/facevalue/vgg-vd-face-fer.mat">
117+
vgg-vd-face-fer</a></td>
118+
<td>vgg faces</td>
119+
<td>fer</td>
120+
<td><a href="http://www.robots.ox.ac.uk/~vgg/data/facevalue/vgg-vd-face-fer.mat">
121+
vgg-vd-face-sfew</a></td>
122+
<td>vgg faces, fer</td>
123+
<td>sfew</td>
124+
</tr>
125+
<tr>
126+
<td><a href="http://www.robots.ox.ac.uk/~vgg/data/facevalue/resnet50-face-bn.mat">
127+
resnet-50-face-fer</a></td>
128+
<td>vgg faces</td>
129+
<td>fer</td>
130+
<td><a href="http://www.robots.ox.ac.uk/~vgg/data/facevalue/resnet50-face-fer-bn.mat">
131+
resnet-50-face-sfew</a></td>
132+
<td>vgg faces, fer</td>
133+
<td>sfew</td>
134+
</tr>
135+
</tbody>
136+
</table>
137+
138+
<p>
139+
The code used to train the models can be found
140+
<a href="https://github.com/albanie/LearningGrimacesByWatchingTV">here</a>.
141+
The faceValue dataset described in the paper is also available for download
142+
<a href="http://www.robots.ox.ac.uk/~vgg/data/facevalue/">here</a>.
143+
</p>
144+
<p>
145+
If you find these models useful, please consider
146+
citing the following paper:
147+
</p>
148+
149+
<em class="citation">
150+
Albanie, Samuel, and Vedaldi, Andrea, "Learning Grimaces by Watching TV."
151+
British Machine Vision Conference (BMVC), 2016.
152+
</em>
153+
</p>
154+
155+
156+
<h5 id="ssd">SSD Detector for matconvnet</h5>
157+
158+
<p>We have created a <a href="https://github.com/albanie/mcnSSD">matconvnet
159+
implementation</a> of the Single Shot Multibox
160+
Detector. The SSD detector, developed by
161+
<a href="https://link.springer.com/chapter/10.1007/978-3-319-46448-0_2">
162+
Wei Liu et al.</a> and published at ECCV 2016) is a powerful framework
163+
for object detection which has demonstrated impressive accuracy at high
164+
frame rates. </p>
165+
166+
<p>Several pretrained caffe models have been made publicly avaiable by
167+
Wei Liu (the code and the links to the models can be found
168+
<a href="https://github.com/weiliu89/caffe/tree/ssd#models">here</a>).
169+
These have been converted for use in matconvnet and can be downloaded
170+
from the links below. In addition, sample models which have been trained
171+
with the matconvnet implementation directly are also available for
172+
download.</p>
173+
174+
</div>
175+
</div>
176+
177+
178+
<div class="row">
179+
<div class="twelve columns">
180+
<table class="pretrained-models">
181+
<thead>
182+
<tr>
183+
<th>model</th>
184+
<th>training data</th>
185+
<th>image size</th>
186+
<th>trained with</th>
187+
<th>Test mAP 2007</th>
188+
</tr>
189+
</thead>
190+
<tbody>
191+
<tr>
192+
<td><a href="http://www.robots.ox.ac.uk/~albanie/models/ssd/ssd-mcn-pascal-vggvd-300.mat">
193+
ssd-mcn-pascal-vggvd-300</a></td>
194+
<td>pascal 2007 trainval + 2012 trainval</td>
195+
<td>300</td>
196+
<td>matconvnet</td>
197+
<td>77.7 (*)</td>
198+
</tr>
199+
<tr>
200+
<td><a href="http://www.robots.ox.ac.uk/~albanie/models/ssd/ssd-mcn-pascal-vggvd-512.mat">
201+
ssd-mcn-pascal-vggvd-512</a></td>
202+
<td>pascal 2007 trainval + 2012 trainval</td>
203+
<td>512</td>
204+
<td>matconvnet</td>
205+
<td>-</td>
206+
</tr>
207+
<tr>
208+
<td><a href="http://www.robots.ox.ac.uk/~albanie/models/ssd/ssd-pascal-vggvd-300.mat">
209+
ssd-pascal-vggvd-300</a></td>
210+
<td>pascal 2007 trainval + 2012 trainval</td>
211+
<td>300</td>
212+
<td>caffe</td>
213+
<td>77.5(*)</td>
214+
</tr>
215+
<tr>
216+
<td><a href="http://www.robots.ox.ac.uk/~albanie/models/ssd/ssd-pascal-vggvd-512.mat">
217+
ssd-pascal-vggvd-512</a></td>
218+
<td>pascal 2007 trainval + 2012 trainval</td>
219+
<td>512</td>
220+
<td>caffe</td>
221+
<td>-</td>
222+
</tr>
223+
<tr>
224+
<td><a href="http://www.robots.ox.ac.uk/~albanie/models/ssd/ssd-pascal-vggvd-ft-300.mat">
225+
ssd-pascal-vggvd-ft-300</a></td>
226+
<td>pascal 2007 trainval + 2012 trainval</td>
227+
<td>300</td>
228+
<td>caffe</td>
229+
<td>77.5(*)</td>
230+
</tr>
231+
<tr>
232+
<td><a href="http://www.robots.ox.ac.uk/~albanie/models/ssd/ssd-pascal-vggvd-ft-512.mat">
233+
ssd-pascal-vggvd-ft-512</a></td>
234+
<td>pascal 2007 trainval + 2012 trainval</td>
235+
<td>512</td>
236+
<td>caffe</td>
237+
<td>-</td>
238+
</tr>
239+
<tr>
240+
<td><a href="http://www.robots.ox.ac.uk/~albanie/models/ssd/ssd-pascal-plus-vggvd-300.mat">
241+
ssd-pascal-plus-vggvd-300</a></td>
242+
<td>pascal 2007 (trainval + test) + 2012 trainval</td>
243+
<td>300</td>
244+
<td>caffe</td>
245+
<td>NA</td>
246+
</tr>
247+
<tr>
248+
<td><a href="http://www.robots.ox.ac.uk/~albanie/models/ssd/ssd-pascal-plus-vggvd-512.mat">
249+
ssd-pascal-plus-vggvd-512</a></td>
250+
<td>pascal 2007 (trainval + test) + 2012 trainval</td>
251+
<td>512</td>
252+
<td>caffe</td>
253+
<td>NA</td>
254+
</tr>
255+
<tr>
256+
<td><a href="http://www.robots.ox.ac.uk/~albanie/models/ssd/ssd-pascal-plus-vggvd-ft-300.mat">
257+
ssd-pascal-plus-vggvd-ft-300</a></td>
258+
<td>pascal 2007 (trainval + test) + 2012 trainval</td>
259+
<td>300</td>
260+
<td>caffe</td>
261+
<td>NA</td>
262+
</tr>
263+
<tr>
264+
<td><a href="http://www.robots.ox.ac.uk/~albanie/models/ssd/ssd-pascal-plus-vggvd-ft-512.mat">
265+
ssd-pascal-plus-vggvd-ft-512</a></td>
266+
<td>pascal 2007 (trainval + test) + 2012 trainval</td>
267+
<td>512</td>
268+
<td>caffe</td>
269+
<td>NA</td>
270+
</tr>
271+
<tr>
272+
<td><a href="http://www.robots.ox.ac.uk/~albanie/models/ssd/ssd-pascal-vggvd-mscoco-300.mat">
273+
ssd-pascal-vggvd-mscoco-300</a></td>
274+
<td>pascal 2007 trainval + 2012 trainval + mscoco trainval 35k</td>
275+
<td>300</td>
276+
<td>caffe</td>
277+
<td>-</td>
278+
</tr>
279+
<tr>
280+
<td><a href="http://www.robots.ox.ac.uk/~albanie/models/ssd/ssd-pascal-vggvd-mscoco-512.mat">
281+
ssd-pascal-vggvd-mscoco-512</a></td>
282+
<td>pascal 2007 trainval + 2012 trainval + mscoco trainval 35k</td>
283+
<td>512</td>
284+
<td>caffe</td>
285+
<td>-</td>
286+
</tr>
287+
</tbody>
288+
</table>
289+
</div>
290+
</div>
291+
292+
<div class="row">
293+
<div class="ten columns">
294+
295+
<p>The following imported models have been trained for evaluation on other
296+
datasets (ilvsrc and ms coco):
297+
</p>
298+
299+
<table class="pretrained-models">
300+
<thead>
301+
<tr>
302+
<th>model</th>
303+
<th>training data</th>
304+
<th>image size</th>
305+
<th>trained with</th>
306+
</tr>
307+
<tr>
308+
<td><a href="http://www.robots.ox.ac.uk/~albanie/models/ssd/ssd-ilsvrc16-vggvd-300.mat">
309+
ssd-ilsvrc16-vggvd-300</a></td>
310+
<td>ilsvrc-16 trainval1</td>
311+
<td>300</td>
312+
<td>caffe</td>
313+
</tr>
314+
<tr>
315+
<td><a href="http://www.robots.ox.ac.uk/~albanie/models/ssd/ssd-ilsvrc15-vggvd-500.mat">
316+
ssd-ilsvrc15-vggvd-500</a></td>
317+
<td>ilsvrc-15 trainval1</td>
318+
<td>500</td>
319+
<td>caffe</td>
320+
</tr>
321+
<tr>
322+
<td><a href="http://www.robots.ox.ac.uk/~albanie/models/ssd/ssd-mscoco-vggvd-300.mat">
323+
ssd-mscoco-vggvd-300</a></td>
324+
<td>mscoco trainval 35k</td>
325+
<td>300</td>
326+
<td>caffe</td>
327+
</tr>
328+
<tr>
329+
<td><a href="http://www.robots.ox.ac.uk/~albanie/models/ssd/ssd-mscoco-vggvd-512.mat">
330+
ssd-mscoco-vggvd-512</a></td>
331+
<td>mscoco trainval 35k</td>
332+
<td>512</td>
333+
<td>caffe</td>
334+
</tr>
335+
</thead>
336+
</tbody>
337+
</table>
338+
339+
<p>Finally, the following atrous VGG-16 model (which was introduced in the
340+
<a href="https://arxiv.org/abs/1412.7062">DeepLab</a> paper) serves as a
341+
good pretrained model to initialise SSD training:
342+
</p>
343+
344+
<table class="pretrained-models">
345+
<thead>
346+
<tr>
347+
<th>model</th>
348+
<th>training data</th>
349+
</tr>
350+
<tr>
351+
<td><a href="http://www.robots.ox.ac.uk/~albanie/models/ssd/vgg-vd-16-reduced.mat"
352+
>vgg-vd-16-reduced</a></td>
353+
<td>ilsvrc</td>
354+
</tr>
355+
</thead>
356+
</tbody>
357+
</table>
358+
359+
360+
<p>If you find this work useful, please consider citing the original SSD
361+
paper:
362+
</p>
363+
<em class="citation">Liu, Wei, et al. "SSD: Single shot multibox detector."
364+
European Conference on Computer Vision.
365+
Springer International Publishing, 2016.</em>
366+
</p>
367+
368+
<div>
369+
<footer>
370+
CSS by <a href="https://github.com/dhg">Mr Dave Gamache</a>
371+
<footer>
372+
</div>
373+
374+
</body>
375+
</html>

‎research.html

+6-1
Original file line numberDiff line numberDiff line change
@@ -58,6 +58,9 @@ <h5 id="subheader">distinctly average computer vision researcher</h5>
5858
<li class="navbar-item">
5959
<a class="navbar-link" href="research.html">Research</a>
6060
</li>
61+
<li class="navbar-item">
62+
<a class="navbar-link" href="models.html">Models</a>
63+
</li>
6164
<li class="navbar-item">
6265
<a class="navbar-link" href="curriculum-vitals.html">CV</a>
6366
</li>
@@ -121,7 +124,9 @@ <h5 style="margin-top: 15%">Projects </h5>
121124
<p>Winner of the <em>award for excellence in
122125
coincidentally/blindly writing, creating titles and submitting
123126
similarly/relatedly/with related titles</em> at SIGBOVIK 2017.</p>
124-
<p>This paper was co-manufactured with the Sébastien Ehrhardt and
127+
<p>This paper was co-manufactured with
128+
<a href="https://fr.linkedin.com/in/sebastien-ehrhardt-85932261">
129+
Sébastien Ehrhardt</a> and
125130
<a href="http://www.robots.ox.ac.uk/~joao/"/>João F. Henriques.</a>.
126131
Code to reproduce the experiments described in the paper can be
127132
found <a href="https://github.com/albanie/SIGBOVIK17-GUNs">here</a>.

0 commit comments

Comments
 (0)
Please sign in to comment.