Deep Unsupervised Image Hashing by Maximizing Bit Entropy (2020-12-22T00:00:00.000000Z)

TL;DR

This work proposes an unsupervised deep hashing layer called Bi-Half Net that maximizes entropy of the binary codes and designs a new parameter-free network layer to explicitly force continuous image features to approximate the optimal half-half bit distribution.

Abstract

Unsupervised hashing is important for indexing huge image or video collections without having expensive annotations available. Hashing aims to learn short binary codes for compact storage and efficient semantic retrieval. We propose an unsupervised deep hashing layer called Bi-Half Net that maximizes entropy of the binary codes. Entropy is maximal when both possible values of the bit are uniformly (half-half) distributed. To maximize bit entropy, we do not add a term to the loss function as this is difficult to optimize and tune. Instead, we design a new parameter-free network layer to explicitly force continuous image features to approximate the optimal half-half bit distribution. This layer is shown to minimize a penalized term of the Wasserstein distance between the learned continuous image features and the optimal half-half bit distribution. Experimental results on the image datasets FLICKR25K, NUS-WIDE, CIFAR-10, MS COCO, MNIST and the video datasets UCF-101 and HMDB-51 show that our approach leads to compact codes and compares favorably to the current state-of-the-art.

Authors

J. V. Gemert

6 papers

Yun-qiang Li

1 papers

TL;DR

Abstract

Authors

References68 items

Push for Quantization: Deep Fisher Hashing

DistillHash: Unsupervised Deep Hashing by Distilling Data Pairs

Unsupervised Deep Hashing with Similarity-Adaptive and Discrete Optimization

Relaxation-Free Deep Hashing via Policy Gradient

Learning Deep Unsupervised Binary Codes for Image Retrieval

Semantic Structure-based Unsupervised Deep Hashing

Deep Hashing via Discrepancy Minimization

Unsupervised Deep Generative Adversarial Hashing Network

Deep Cauchy Hashing for Hamming Space Retrieval

MINE: Mutual Information Neural Estimation

Can Spatiotemporal 3D CNNs Retrace the History of 2D CNNs and ImageNet?

Learning Spatio-Temporal Features with 3D Residual Networks for Action Recognition

Learning Multifunctional Binary Codes for Both Category and Attribute Oriented Retrieval Tasks

Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour

Deep Supervised Discrete Hashing

The Kinetics Human Action Video Dataset

Deep Hashing for Scalable Image Search

Simultaneous Feature Aggregating and Hashing for Large-Scale Image Search

HashNet: Deep Learning to Hash by Continuation

Stochastic Generative Hashing

Deep Supervised Hashing for Fast Image Retrieval

Learning Compact Binary Descriptors with Unsupervised Deep Neural Networks

InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets

Deep Cross-Modal Hashing

Deep Residual Learning for Image Recognition

Scalable Graph Hashing with Feature Transformation

Deep hashing for compact binary codes learning

Simultaneous feature learning and hash coding with deep neural networks

Supervised Discrete Hashing

Deep semantic ranking based hashing for multi-label image retrieval

Optimizing affinity-based binary hashing using auxiliary coordinates

Discrete Graph Hashing

Very Deep Convolutional Networks for Large-Scale Image Recognition

Supervised hashing with latent factor models

Principal Component Analysis (PCA)

Microsoft COCO: Common Objects in Context

One weird trick for parallelizing convolutional neural networks

Fast Supervised Hashing with Decision Trees for High-Dimensional Data

Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation

K-Means Hashing: An Affinity-Preserving Quantization Method for Learning Binary Compact Codes

Isotropic Hashing

UCF101: A Dataset of 101 Human Actions Classes From Videos in The Wild

A new approach to interdomain routing based on secure multi-party computation

Supervised hashing with kernels

HMDB: A large video database for human motion recognition

Hashing with Graphs

Iterative quantization: A procrustean approach to learning binary codes

Non-Metric Locality-Sensitive Hashing

Self-taught hashing for fast similarity search

Locality-sensitive binary codes from shift-invariant kernels

Fast Similarity Search for Learned Metrics

Kernelized locality-sensitive hashing for scalable image search

NUS-WIDE: a real-world web image database from National University of Singapore

The MIR flickr retrieval evaluation

Hamming Embedding and Weak Geometric Consistency for Large Scale Image Search

Elements of Information Theory

Locality-sensitive hashing scheme based on p-stable distributions

Topics in Optimal Transportation

Similarity Search in High Dimensions via Hashing

Greedy Hash: Towards Fast Optimization for Accurate Hash Coding in CNN

Raziperchikolaei and Carreira-Perpi ˜ n ´ an

and Lazebnik

Learning Multiple Layers of Features from Tiny Images

and Grauman

A Mathematical Theory of Communication

and Indyk

Gradient-based learning applied to document recognition

Proceedings of the Twenty-Third International Joint Conference on Artificial Intelligence Harmonious Hashing

Field of Study

Venue Information

Name

Type

URL

Alternate Names