Deep Clustering for Unsupervised Learning of Visual Features

Abstract

Clustering is a class of unsupervised learning methods that has been extensively applied and studied in computer vision. Little work has been done to adapt it to the end-to-end training of visual features on large scale datasets. In this work, we present DeepCluster, a clustering method that jointly learns the parameters of a neural network and the cluster assignments of the resulting features. DeepCluster iteratively groups the features with a standard clustering algorithm, k-means, and uses the subsequent assignments as supervision to update the weights of the network. We apply DeepCluster to the unsupervised training of convolutional neural networks on large datasets like ImageNet and YFCC100M. The resulting model outperforms the current state of the art by a significant margin on all the standard benchmarks.

Accepted at ECCV 2018

References in corpus (16)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
• Densely Connected Convolutional Networks
• Understanding Neural Networks Through Deep Visualization
• Adversarial Feature Learning
• Particular object retrieval with integral max-pooling of CNN activations
• Adversarially Learned Inference
• Unsupervised Learning by Predicting Noise
• Optimizing the Latent Space of Generative Networks
• Data-dependent Initializations of Convolutional Neural Networks
• Convolutional Kernel Networks
• Transitive Invariance for Self-supervised Visual Representation Learning
• ConvNets and ImageNet Beyond Accuracy: Understanding Mistakes and Uncovering Biases
• Split-Brain Autoencoders: Unsupervised Learning by Cross-Channel Prediction
• An evaluation of large-scale methods for image instance and class discovery

Cited by in corpus (69)

• Barlow Twins: Self-Supervised Learning via Redundancy Reduction
• Graph Self-Supervised Learning: A Survey
• Neural Image Compression for Gigapixel Histopathology Image Analysis
• SPICE: Semantic Pseudo-labeling for Image Clustering
• Unsupervised Point Cloud Representation Learning with Deep Neural Networks: A Survey
• Twin Contrastive Learning for Online Clustering
• Mixing Up Contrastive Learning: Self-Supervised Representation Learning for Time Series
• Self-Supervised Multisensor Change Detection
• Understanding Dimensional Collapse in Contrastive Self-supervised Learning
• Self-Supervised Representation Learning for Astronomical Images
• Multi-level Second-order Few-shot Learning
• AET vs. AED: Unsupervised Representation Learning by Auto-Encoding Transformations rather than Data
• Self-supervised Consensus Representation Learning for Attributed Graph
• Deep Discriminative Clustering Analysis
• Identity-Guided Human Semantic Parsing for Person Re-Identification
• Assume, Augment and Learn: Unsupervised Few-Shot Meta-Learning via Random Labels and Data Augmentation
• Prior-Knowledge and Attention-based Meta-Learning for Few-Shot Learning
• Pixel-Level Clustering Network for Unsupervised Image Segmentation
• X-GOAL: Multiplex Heterogeneous Graph Prototypical Contrastive Learning
• Sharing Matters for Generalization in Deep Metric Learning
• Adaptive Soft Contrastive Learning
• Image Clustering using an Augmented Generative Adversarial Network and Information Maximization
• Semi-Supervised Visual Representation Learning for Fashion Compatibility
• Self-supervised learning for autonomous vehicles perception: A conciliation between analytical and learning methods
• Learning Discriminative Motion Features Through Detection
• Deep Unsupervised Learning for 3D ALS Point Cloud Change Detection
• Towards Cross-Granularity Few-Shot Learning: Coarse-to-Fine Pseudo-Labeling with Visual-Semantic Meta-Embedding
• Learning Group Activities from Skeletons without Individual Action Labels
• Beyond Cats and Dogs: Semi-supervised Classification of fuzzy labels with overclustering
• Large-Scale Hyperspectral Image Clustering Using Contrastive Learning
• Fuzzy Overclustering: Semi-Supervised Classification of Fuzzy Labels with Overclustering and Inverse Cross-Entropy
• More for Less: Non-Intrusive Speech Quality Assessment with Limited Annotations
• Modeling Multiple Views via Implicitly Preserving Global Consistency and Local Complementarity
• Self-supervising Fine-grained Region Similarities for Large-scale Image Localization
• Y-net: Biomedical Image Segmentation and Clustering
• Distilling Localization for Self-Supervised Representation Learning
• Explaining Deep Neural Networks using Unsupervised Clustering
• Fetal Gender Identification using Machine and Deep Learning Algorithms on Phonocardiogram Signals
• It Takes Two to Tango: Mixup for Deep Metric Learning
• Cluster-Guided Unsupervised Domain Adaptation for Deep Speaker Embedding
• CARL-G: Clustering-Accelerated Representation Learning on Graphs
• On The State of Data In Computer Vision: Human Annotations Remain Indispensable for Developing Deep Learning Models
• Representation Learning via Consistent Assignment of Views to Clusters
• KinePose: A temporally optimized inverse kinematics technique for 6DOF human pose estimation with biomechanical constraints
• AVT: Unsupervised Learning of Transformation Equivariant Representations by Autoencoding Variational Transformations
• Self-Supervised Pre-Training Boosts Semantic Scene Segmentation on LiDAR Data
• Bootstrapping Contrastive Learning Enhanced Music Cold-Start Matching
• SUVR: A Search-based Approach to Unsupervised Visual Representation Learning
• Clustering-Based Representation Learning through Output Translation and Its Application to Remote--Sensing Images
• Jigsaw Clustering for Unsupervised Visual Representation Learning
• DISCERN: Diversity-based Selection of Centroids for k-Estimation and Rapid Non-stochastic Clustering
• Parallelization of Kmeans++ using CUDA
• Semi-Supervised Class Discovery
• Improving Image Clustering through Sample Ranking and Its Application to remote--sensing images
• Class Introspection: A Novel Technique for Detecting Unlabeled Subclasses by Leveraging Classifier Explainability Methods
• Reliable Label Bootstrapping for Semi-Supervised Learning
• Deep Transformation-Invariant Clustering
• AETv2: AutoEncoding Transformations for Self-Supervised Representation Learning by Minimizing Geodesic Distances in Lie Groups
• Neural Rejuvenation: Improving Deep Network Training by Enhancing Computational Resource Utilization
• clusterNOR: A NUMA-Optimized Clustering Framework
• On the robustness of self-supervised representations for multi-view object classification
• A neuro-inspired architecture for unsupervised continual learning based on online clustering and hierarchical predictive coding
• SERE: Exploring Feature Self-relation for Self-supervised Transformer
• ScaleNet: An Unsupervised Representation Learning Method for Limited Information
• Domain-Agnostic Clustering with Self-Distillation
• A Self-supervised Learning System for Object Detection in Videos Using Random Walks on Graphs
• UCSL : A Machine Learning Expectation-Maximization framework for Unsupervised Clustering driven by Supervised Learning
• Discriminative Clustering with Representation Learning with any Ratio of Labeled to Unlabeled Data
• Constructing Multiple Tasks for Augmentation: Improving Neural Image Classification With K-means Features