British Machine Vision Conference (BMVC), London 2009

What can the world tell us about an image?

author: Alexei Efros, Carnegie Mellon University
published: Dec. 1, 2009, recorded: September 2009, views: 1468

Slides

Slides
0:00	What can we learn from a single image?
0:29	What do we see?
1:41	The Miserable Life of an Object Detector
1:57	What the Detector Sees
2:07	State-of-the-Art Pedestrian DetectionTrue
2:51	Importance of Looking Globally
3:27	Detail on Monet's painting
3:43	Importance of Looking Globally
3:49	Detail on Monet's painting
4:33	Seeing less than you think… (1)
5:29	Seeing less than you think… (2)
6:17	Real Relationships are 3D
6:57	Imaging Process
7:37	Unsolvable Problem
7:53	Ecological Optics
8:21	Our World is Structured
9:17	Understanding Scenes
11:05	Support
11:25	Size
11:49	Interposition
11:53	Position, Probability, Size
12:04	+ illumination
12:24	Our Goal: Scene Understanding
13:30	Ecological Statistics
14:18	Collaborators
15:34	Scene Layout
16:38	Learn from labeled data
16:58	What cues to use?
17:58	Weak Geometric Cues
18:18	Need Good Spatial Support
19:30	Image Segmentation
20:54	Estimating surfaces from segments
21:22	Labeling Segments
21:58	Image Labeling
22:18	Results
22:45	No Hard Decisions
23:29	Indoor Images
23:33	Paintings
23:41	Graphics application:Automatic Photo Pop-up
24:30	- (missing examples of images)
25:32	Failures
25:52	Occlusions are everywhere!
26:45	Finding occlusions
27:53	Occlusion Reasoning as Classification
28:55	Object Size / Camera ViewpointWorld
30:07	Camera viewpoint for LabelMe
32:27	Helping Object Detection (1)
33:19	Helping Object Detection (2)
33:59	More Chickens, More Eggs…
34:30	Best Guesses
35:22	Putting it all together
35:31	Some Results (1)
35:51	Some Results (2)
36:03	More Results
36:35	Putting Objects in Perspective
36:55	Illumination from a Single Image (1)
38:19	Illumination from a Single Image (2)
39:07	Illumination from a Single Image (3)
39:27	Algorithm
40:43	Weak cues
43:27	Data-driven Sun Elevation Prior (1)
44:19	Data-driven Sun Elevation Prior (2)
44:44	Scene Semantics:Understanding the EntireScene
45:16	Hays & Efros, SIGGRAPH‘07 (1)
45:32	Hays & Efros, SIGGRAPH‘07 (2)
45:40	Where does the knowledge come from?
45:56	Scene Semantics!
46:31	Images example
46:54	Scene Completion Result
47:01	Example, Scene Descriptor (1)
47:09	Example, Scene Descriptor (2)
47:12	Example, Scene Descriptor (3)
47:32	Gist scene descriptor (1)
47:56	Gist scene descriptor (2)
48:08	Gist scene descriptor (3)
48:15	... 200 total
48:21	Gist scene descriptor (4)
48:29	Graph cut + Poisson blending (missing picture examples) (1)
48:48	Example, Scene Descriptor (4)
48:56	Example, Scene Descriptor (5)
48:59	Example, Scene Descriptor (6)
49:00	Example, Scene Descriptor, ... 200 scene matches (1)
49:12	Example, Scene Descriptor (7)
49:40	Example, Scene Descriptor (8)
49:43	Example, Scene Descriptor (9)
49:47	Example, Scene Descriptor (10)
49:49	Example, Scene Descriptor (11)
49:53	Example, Scene Descriptor (12)
49:58	Example, Scene Descriptor (13)
50:02	Example, Scene Descriptor (14)
50:03	Example, Scene Descriptor (15)
50:12	Example, Scene Descriptor, ... 200 scene matches (2)
50:31	Example, Scene Descriptor (16)
51:07	Example, Scene Descriptor (17)
51:11	Why does it work?
51:22	Example, Scene Descriptor (18)
51:30	10 nearest neighbors from acollection of 20,000 images (1)
51:54	10 nearest neighbors from acollection of 20,000 images (2)
52:18	Database of 70 Million 32x32 images
52:46	The Big Picture
53:19	im2gps (Hays & Efros, CVPR 2008)
53:51	How much can an image tell about its geographic location?
54:02	Image (and its geographic location), example 1
54:08	(Image and) its geographic location, example (1)
54:32	Im2gps (1)
55:16	Example Scene Matches
55:24	Voting Scheme
55:44	im2gps (2)
56:30	Image (and its geographic location), example 2
56:40	(Image and) its geographic location, example (2)
56:44	Image (and its geographic location), example 3
57:01	(Image and) its geographic location, example (3)
57:25	Image (and its geographic location), example 4
57:26	(Image and) its geographic location, example (4)
57:30	Image (and its geographic location), example 5
57:33	(Image and) its geographic location, example (5)
57:39	Image (and its geographic location), example 5
57:41	(Image and) its geographic location, example (5)
58:02	Image (and its geographic location), example 6
58:06	(Image and) its geographic location, example (6)
58:13	Image (and its geographic location), example 6
58:19	(Image and) its geographic location, example (6)
58:22	Data-driven categories
58:45	Data-driven categories, geographic location
59:01	Data-driven categories, Elevation gradient = 112 m / km*
59:32	Elevation gradient magnitude ranking
59:42	Figure 2, Global population density map
59:46	Population density ranking
59:49	Figure 4, Global land cover classification map
59:59	Barren or sparsely populated
60:00	Urban and built up
60:03	Snow and Ice
60:15	Savannah
60:17	Water
60:19	Conclusions

Report a problem or upload files

If you have found a problem with this lecture or would like to send us extra material, articles, exercises, etc., please use our ticket system to describe your request and upload the data.
Enter your e-mail into the 'Cc' field, and we will keep you updated with your request's status.

Lecture popularity: You need to login to cast your vote.