Главная
Study mode:
on
1
Object recognition
2
Bag of Words
3
Bags of Words
4
Capture the pattern in patch
5
The dimensionality
6
Sample many patches
7
Two types of sampling
8
Sample many images
9
Include all relevant variations
10
Form a dictionary of words
11
Ideally words cover similar patches
12
Count words per image
13
Codeboods
14
Learn histogram similarity
15
Similarity between two histograms
16
Classify unknown image
17
Concept-specific codebooks
18
Conclusion on concept-codebooks
19
Soft word assignment
20
Fisher vector
21
Conclusion on words
22
Codebook synonyms
23
How close are synonyms?
24
90% removed, same result
25
Visual synonym examples
26
Conclusion on visual synonyms
27
Convex reduced codebooks
28
Conclusion convex reduced
29
The where and what
30
What makes a boat a boat?
31
What is the object in the middle?
32
Where is evidence?
33
Where is evidence for an object?
34
The visual extent of an object
35
Context dominance
36
Object dominance
37
Object detail dominance
38
Pyramids: simple compositional
39
Exhaustive search
40
The need for high recall
41
The need for hierarchy
42
Selective search example
43
Selective search to get high recall
44
Average best overlap -88%
45
Classification with selective search
46
Conclusion on location
47
Two concepts in interaction
Description:
Explore object recognition techniques in this guest presentation by Dr. Arnold Smeulders. Delve into Bag of Words models, patch sampling methods, and dictionary formation for image classification. Learn about concept-specific codebooks, soft word assignment, and Fisher vectors. Examine visual synonyms and convex reduced codebooks. Investigate the importance of object location, context, and hierarchical search methods like selective search. Understand the interplay between object classification and localization in computer vision tasks.

Object Recognition

University of Central Florida
Add to list
#Computer Science #Artificial Intelligence #Computer Vision #Object Detection #Machine Learning #Image Processing #Pattern Recognition #Object Recognition #Bag of Words
00:00
-01:25
1 of 47

1 Object recognition