Download presentation

Presentation is loading. Please wait.

Published byTaylor Gerrish Modified over 3 years ago

1
1/26 The Inverted Multi-Index VGG Oxford, 25 Oct 2012 Victor Lempitsky joint work with Artem Babenko

2
2/26 From images to descriptors Set of 128D descriptors Interesting point description: Interesting point detection:

3
3/26 Query process Dataset of visual descriptors Image set: Query: Important extras: + geometric verification + query expansion Main operation: Finding similar descriptors

4
4/26 Demands Initial setup: Dataset size: few million images Typical RAM size: few dozen gigabytes Tolerable query time: few seconds Search problem: Dataset size: few billion features Feature footprint: ~ a dozen bytes Tolerable time: few milliseconds per feature Dataset of visual descriptors Each image has ~1000 descriptors nearest neighbor search problem we are tackling

5
5/26 Meeting the demands Main observation: the vectors have a specific structure: correlated dimensions, natural image statistics, etc… Technologies: Dimensionality reduction Vector quantization Inverted index Locality-sensitive hashing Product quantization Binary/Hamming encodings Best combinations (previous state-of-the-art): Inverted index + Product Quantization [Jegou et al. TPAMI 2011] Inverted index + Binary encoding [Jegou et al. ECCV 2008] Our contribution: Inverted Multi-Index New state-of-the-art for BIGANN: Inverted multi-index + Product Quantization [CVPR 2012]

6
6/26 The inverted index Visual codebook "Visual word" Sivic & Zisserman ICCV 2003

7
7/26 Querying the inverted index Have to consider several words for best accuracy Want to use as big codebook as possible Want to spend as little time as possible for matching to codebooks conflict Query:

8
8/26 Product quantization [Jegou, Douze, Schmid // TPAMI 2011]: 1.Split vector into correlated subvectors 2.use separate small codebook for each chunk For a budget of 4 bytes per descriptor: 1.Can use a single codebook with 1 billion codewords many minutes 128GB 2.Can use 4 different codebooks with 256 codewords each < 1 millisecond 32KB IVFADC+ variants (state-of-the-art for billion scale datasets) = inverted index for indexing + product quantization for reranking Quantization vs. Product quantization:

9
9/26 The inverted multi-index Our idea: use product quantization for indexing Main advantage: For the same K, much finer subdivision achieved Main problem: Very non-uniform entry size distribution

10
10/26 Querying the inverted multi-index 1 1 2 2 3 3 4 4 5 5 6 6 7 7 8 8 9 9 10 Input: query Output: stream of entries Answer to the query:

11
11/26 Querying the inverted multi-index – Step 1

12
12/26 Querying the inverted multi-index – Step 2 0.60.84.16.18.19.1 2.52.7681011 3.53.7791112 6.56.710121415 7.57.711131516 11.511.7 15171920 0.60.84.16.18.19.1 2.52.7681011 3.53.7791112 6.56.710121415 7.57.711131516 11.511.7 15171920 0.60.84.16.18.19.1 2.52.7681011 3.53.7791112 6.56.710121415 7.57.711131516 11.511.7 15171920 0.60.84.16.18.19.1 2.52.7681011 3.53.7791112 6.56.710121415 7.57.711131516 11.511.7 15171920 0.60.84.16.18.19.1 2.52.7681011 3.53.7791112 6.56.710121415 7.57.711131516 11.511.7 15171920 0.60.84.16.18.19.1 2.52.7681011 3.53.7791112 6.56.710121415 7.57.711131516 11.511.7 15171920 123456123456123456123456123456123456 1 2 3 4 5 6 Step 2: the multi-sequence algorithm

13
13/26 Querying the inverted multi-index

14
14/26 Experimental protocol Dataset: 1.1 billion of SIFT vectors [Jegou et al.] 2.Hold-out set of 10000 queries, for which Euclidean nearest neighbors are known Comparing index and multi-index: Set a candidate set length T For each query: Retrieve closest entries from index or multi-index and concatenate lists Stop when the next entry does not fit For small T inverted index can return empty list Check whether the true neighbor is in the list Report the share of queries where the neighbor was present (recall@T)

15
15/26 Performance comparison Recall on the dataset of 1 billion of visual descriptors: 100 x Time increase: 1.4 msec -> 2.2 msec on a single core (with BLAS instructions) "How fast can we catch the nearest neighbor to the query?" K = 2 14

16
16/26 Performance comparison Recall on the dataset of 1 billion 128D visual descriptors:

17
17/26 Time complexity For same K index gets a slight advantage because of BLAS instructions

18
18/26 Memory organization Overhead from multi-index: Averaging over N descriptors:

19
19/26 Why two? For larger number of parts: Memory overhead becomes larger Population densities become even more non-uniform (multi-sequence algorithm has to work harder to accumulate the candidates) In our experiments, 4 parts with small K=128 may be competitive for some datasets and reasonably short candidate lists (e.g. duplicate search). Indexing is blazingly fast in these cases!

20
20/26 Multi-Index + Reranking "Multi-ADC": use m bytes to encode the original vector using product quantization "Multi-D-ADC": use m bytes to encode the remainder between the original vector and the centroid Same architecture as IVFADC of Jegou et al., but replaces index with multi-index faster (efficient caching possible for distance computation) more accurate Evaluation protocol: 1.Query the inverted index for T candidates 2.Reconstruct the original points and rerank according to the distance to the query 3.Look whether the true nearest neighbor is within top T*

21
21/26 Multi-ADC vs. Exhaustive search

22
22/26 Multi-D-ADC vs State-of-the-art State-of-the-art [Jegou et al.] Combining multi-index + reranking:

23
23/26 Performance on 80 million GISTs Multi-D-ADC performance: Index vs Multi-index: Same protocols as before, but on 80 million GISTs (384 dimensions) of Tiny Images [Torralba et al. PAMI'08]

24
24/26 Retrieval examples Exact NN Uncompressed GIST Multi-D-ADC 16 bytes Exact NN Uncompressed GIST Multi-D-ADC 16 bytes Exact NN Uncompressed GIST Multi-D-ADC 16 bytes Exact NN Uncompressed GIST Multi-D-ADC 16 bytes

25
25/26 Multi-Index and PCA (128->32 dimensions)

26
26/26 Conclusions A new data structure for indexing the visual descriptors Significant accuracy boost over the inverted index at the cost of the small memory overhead Code available (will soon be online)

27
27/26 Other usage scenarios Large-scale NN search' based approaches: Holistic high dimensional image descriptors: GISTs, VLADs, Fisher vectors, classemes… Pose descriptors Other multimedia Additive norms and kernels: L1, Hamming, Mahalanobis, chi-square kernel, intersection kernel, etc. (Mostly) straightforward extensions possible:

28
28/26 Visual search Van Gogh, 1890 "Landscape with Carriage and Train in the Background" Pushkin museum, Moscow Collection of many millions of fine art images The closest match: What is this painting? Learn more about it

Similar presentations

Presentation is loading. Please wait....

OK

Adding Up In Chunks.

Adding Up In Chunks.

© 2018 SlidePlayer.com Inc.

All rights reserved.

To make this website work, we log user data and share it with processors. To use this website, you must agree to our Privacy Policy, including cookie policy.

Ads by Google

Ppt on 7 wonders of the world 2012 Ppt on series and parallel combination of resistors Ppt on eco friendly agricultural practices Ppt on swami vivekananda How to open ppt on mac Ppt on modern science and technology Ppt on number system for class 5 Ppt on sports day japan Ppt on life of william shakespeare Ppt on data collection methods quantitative research