
en
0.25
0.5
0.75
1.25
1.5
1.75
2
The quest for a theory of vision: from the level framework (revised) to the invariances of the ventral stream
Published on Feb 4, 20256097 Views
Related categories
Presentation
The quest for a theory of vision: from the level framework (revised) to the invariances of the ventral stream00:00
Collaborators in recent work67:30:09
Vision as Intelligence74:20:47
Vision: a very difficult computational problem92:49:52
The problem of intelligence (in particular, vision)99:44:30
Deep Blue109:11:30
Watson115:35:15
Siri118:15:16
Drones120:19:04
Vision @CBCL, ~20 years ago - 1132:45:13
Vision @CBCL, ~20 years ago - 2134:41:02
Vision @CBCL, ~20 years ago - 3137:30:25
Vision @CBCL, ~18 years ago - 1140:02:12
Vision @CBCL, ~18 years ago - 2143:03:09
Vision @CBCL, ~18 years ago - 3143:17:45
Vision @CBCL, ~18 years ago - 4143:56:54
Vision, ~ now148:22:48
MobilEye - 1149:56:07
MobilEye - 2152:56:07
Advertisement by Volvo158:38:23
Recent successes in ML and AI: computer vision172:59:28
A "fullTuring" test for vision? - 1180:32:04
A "fullTuring" test for vision? - 2182:24:36
A "fullTuring" test for vision? - 3184:32:45
A "fullTuring" test for vision? - 4185:48:29
A "fullTuring" test for vision? - 5186:38:02
A "fullTuring" test for vision? - 6187:04:16
A "fullTuring" test for vision? - 7187:24:39
A "fullTuring" test for vision? - 8188:27:16
A "fullTuring" test for vision? - 9188:59:16
A "fullTuring" test for vision? - 10194:53:12
The next big step: brain209:54:25
Vision in the Brain217:45:49
Visual Object Recognition: the ventral stream (macaque)229:42:01
Recognition in the Ventral Stream: "standard" feedforward model266:38:34
Recognition in Visual Cortex: "classical model", selective and invariant283:35:59
Model "works": it accounts for physiology293:17:13
Model "works": it accounts for psychophysics - 1302:25:55
Model "works": it accounts for psychophysics - 2302:39:30
Model "works": it accounts for psychophysics - 3309:16:02
Model "works": it performs well at computational level - 1320:48:22
Model "works": it performs well at computational level - 2321:10:12
Model "works": it performs well at computational level - 3326:32:57
Model "works": it performs well at computational level - 4329:57:59
Model "works": it performs well at computational level - 5330:16:57
Visual Cortex: models and theories341:20:59
A theory (unpublished) of the ventral stream: too nice to be true?360:02:12
Motivation: transformations may be a main difficulty for (biological) object recognition393:36:26
Motivation: transformations may be the main difficulty for (biological) object recognition404:13:27
Conjecture: the key computational problem "solved" by the ventral stream414:47:24
A theory of invariant recognition in the ventral stream421:38:29
Some of the questions answered by the theory433:27:38
Key theorems456:04:23
Gabor-like tuning with "universal constants" in simple cells463:20:47
2 Different stages in the theory492:33:27
Image representation in the ventral stream512:11:57
Templates and signature528:33:34
A motivation for signatures: the Johnson-Lindenstrauss theorem (features do not matter much!)540:48:50
Geometric transformations560:20:37
Learning to be invariant for any new object566:35:35
Templatebook595:16:47
Invariance lemma600:27:36
Invariant Signature609:49:23
One-layer architectures: pro and contra615:11:30
Whole-parts invariance624:29:52
Whole-parts theorem640:44:54
Part II: Linking Conjecture644:39:08
Tuning and eigenvectors of covariance matrix672:09:52
Gaussian aperture: the cortical equation678:30:16
The cortical equation: general properties701:23:06
Cortical equation in 2D: natural images, Gabor-like receptive fields - 1710:02:42
Cortical equation in 2D: natural images, Gabor-like receptive fields - 2717:20:52
Too nice to be true!726:01:17
Beyond V1, towards V2 and V4735:47:18
We are working on implementing the full theory744:56:37
Class-specific modules756:51:49
A theory of biological vision765:17:00