Similar image search
- 1. Similar Image Search with a Tiny Bag-of-Delegates Representation Weiwen Tu, Rong Pan, Jingdong Wang
- 2. Similar Image Search • text-based image search • context-based image search
- 6. Example for Tiny Bag-of-Delegates
- 8. Spatial Partition Tree • Spatial partition tree is a method for recursively subdividing a space into subsets by hyperplanes. • Popular data structures: Kd-tree, PCA-tree, VP-tree.
- 9. Example for Spatial Partition Tree The left figure shows a spatial partitioning induced by an RP tree. The cross is q query point and the lines means the partition hyperplanes.
- 11. Problem A larger amount of trees yield a better performance, while requiring more storage to save the inverted indices and accordingly more query time Problem: Can we use a small amount of trees but with the search performance guaranteed?
- 13. Criterion
- 14. Criterion
- 17. Random selection vs. forward selection Random selection Forward selection
- 18. Adaptive scheme
- 19. Adaptive scheme
- 20. Adaptive scheme
- 21. Adaptive forward selection 1. Initialization: 𝑄 ← 𝐿, 𝑡 ← 0, 𝑒 ← 𝑄 , 𝑅 ← ∅ 2. Repeat 3. Candidate proposal Randomly generate spatial partition tree 𝒯 per Q 4. Candidate selection Choose the spatial partition tree T from 𝒯 that keeps the large number of Q 5. Update Discard all the pairs of points lying in the same bucket in T from Q 𝑡 ← 𝑡 + 1, 𝑒 ← 𝑄 , 𝑅 ← 𝑅 ∪ {𝑇} 6. Until 𝑒 ≤ 𝜀 && 𝑡 ≥ 𝜏 7. Return R;
- 22. EXPERIMENTS • Data set – 32 ×32 color images – 1M tiny images to form around 80M images – global GIST descriptor(384 D vector) • Evaluation criteria(average accuracy score)
- 23. Evaluation
- 24. Accuracy vs indices Bucket size: the maximum number of points in a leaf node; g: the number of target NNs.
- 25. Accuracy vs indices bucket size = 300 bucket size = 100
- 26. accuracy vs accessed images
- 28. Visual search result 13 : 9 22 : 9
- 29. Thank you!