CV Chapter 5 Local Features and Matching
Questions about the lecture 'Computer Vision' of the RWTH Aachen Chapter 5 Local Features and Matching
Questions about the lecture 'Computer Vision' of the RWTH Aachen Chapter 5 Local Features and Matching
Set of flashcards Details
| Flashcards | 55 |
|---|---|
| Language | English |
| Category | Computer Science |
| Level | University |
| Created / Updated | 04.02.2017 / 20.02.2017 |
| Weblink |
https://card2brain.ch/box/20170204_cv_chapter_5_local_features_and_matching
|
| Embed |
<iframe src="https://card2brain.ch/box/20170204_cv_chapter_5_local_features_and_matching/embed" width="780" height="150" scrolling="no" frameborder="0"></iframe>
|
What are two techniques to deal with outliers for recognition with local features? [2]
1. Random sample consensus (RANSAC) and 2. generalized hough transform
Define the RANSAC algorithm? [5]
1. Randomly select a seed group
2. Compute transformation from seed group
3. Find inlier to this transformation within threshold
4. Recompute least-squares estimate of transformation
5. Keep transformations with largest number of inlier
What are the probabilities for RANSAC? [2]
1. Probability that 1 sample of n points is correct: w^n with w fraction of inlier
2. Probability that k samples fail: (1-w^n)^k
What is a currently available application for indexing with local features?
Google goggles
What are the characteristic of Google goggles? [2]
1. Mobile visual search
2. Large-scale image matching problem
What is the parallel between indexing with local features and text retrieval? [2]
1. Find all pages on which a word occurs
2. Find all images in which a feature occurs
Name two solutions for indexing with local features? [2]
1. Inverted index files
2. Bag of words
What is the principle of inverted index files? [2]
1. Index features as point in high dimensional feature space
2. Search for close points in feature space
What is the complexity of search in feature space for inverted index files? [2]
1. Efficient for low-dimensional descriptors
2. Approximate for high-dimensional descriptors
How can the feature space be structured for inverted index files? [2]
1. Quantize feature space via clustering and defining prototype words
2. Vocabulary tree filled via training
What is the definition of term frequency - inverse document frequency (tf-idf) for inverted index files? [3]
1. Standard for text retrieval
2. Describe frame by frequencies of features within
3. t_i = n_id/n_d *log(N/n_i)
n_i number of occurrences of i in database
n_id number of occurrences of i in document d
n_d number of occurrences in document d
N total number of documents in database
What is the principle of bag of words? [3]
1. Split several images into visual words removing spatial layout
2. For every image specify a histogram of all words // Vector
3. Compare histograms with any known measures
To which technique is bag of words similar to?
Similar to bag of textons for texture representation
What can be done to save the spatial layout in bag of words? [2]
1. Spatial pyramid representation
2. Perform BoW for each sub-gird of an image
How is the performance of bag of words compared to the parts and shape model?
Better
