Skip to main content

Table 2 Image search results on the CIFAR-10 dataset for different deep learning hashing methods

From: Deep learning hashing for mobile visual search

ID Method 12/16-bits 32-bits 48-bits 64-bits
1 Deep learning of binary hash [68] 89.30% 89.72% 89.73% -
2 Deep hashing for compact binary codes [69] 46.75% 51.01% - 52.50%
3 Unsupervised deep neural networks hashing [23] 19.43% 24.86% - 27.73%
4 Supervised deep hashing [22] 46.5% 52.1% 53.2% -
5 Semantics-preserving deep hashing [23] - - 89.97% -
6 Binary deep neural network hashing [26] 67.32% 69.62% - -
7 Bit-scalable deep hashing [27] 55.2% 55.8% 58.1% -
8 One-stage deep hashing [28] 61.46% 62.87% 63.05% 63.26%
9 Deep pairwise-supervised hashing [29] 71.3% 74.4% 75.7% -
10 Deep hashing network [72] 55.5% 60.3% 62.1% -
11 Deep semantic-preserving and ranking-based hashing [73] ≈ 78% ≈ 78% ≈ 77% -
  1. The accuracy in terms of MAP with different hash bits length, which are collected from their papers