Unlabeled Samples Generated by GAN Improve the Person Re-identification Baseline in vitro

Abstract

The main contribution of this paper is a simple semi-supervised pipeline that only uses the original training set without collecting extra data. It is challenging in 1) how to obtain more training data only from the training set and 2) how to use the newly generated data. In this work, the generative adversarial network (GAN) is used to generate unlabeled samples. We propose the label smoothing regularization for outliers (LSRO). This method assigns a uniform label distribution to the unlabeled images, which regularizes the supervised model and improves the baseline. We verify the proposed method on a practical problem: person re-identification (re-ID). This task aims to retrieve a query person from other cameras. We adopt the deep convolutional generative adversarial network (DCGAN) for sample generation, and a baseline convolutional neural network (CNN) for representation learning. Experiments show that adding the GAN-generated data effectively improves the discriminative ability of learned CNN embeddings. On three large-scale datasets, Market-1501, CUHK03 and DukeMTMC-reID, we obtain +4.37%, +1.6% and +2.46% improvement in rank-1 precision over the baseline CNN, respectively. We additionally apply the proposed method to fine-grained bird recognition and achieve a +0.6% improvement over a strong baseline. The code is available at https://github.com/layumi/Person-reID_GAN.

9 pages, 6 figures, accepted by ICCV 2017

References in corpus (6)

• Learning a Probabilistic Latent Space of Object Shapes via 3D Generative-Adversarial Modeling
• Person Re-identification: Past, Present and Future
• Deep Transfer Learning for Person Re-identification
• Looking Beyond Appearances: Synthetic Training Data for Deep CNNs in Re-identification
• Performance Measures and a Data Set for Multi-Target, Multi-Camera Tracking
• Deep Metric Learning for Practical Person Re-Identification

Cited by in corpus (66)

• Random Erasing Data Augmentation
• SphereReID: Deep Hypersphere Manifold Embedding for Person Re-Identification
• Margin Sample Mining Loss: A Deep Learning Based Method for Person Re-identification
• Beyond Face Rotation: Global and Local Perception GAN for Photorealistic and Identity Preserving Frontal View Synthesis
• SVDNet for Pedestrian Retrieval
• Training Confidence-calibrated Classifiers for Detecting Out-of-Distribution Samples
• The Devil is in the Middle: Exploiting Mid-level Representations for Cross-Domain Instance Matching
• Pedestrian-Synthesis-GAN: Generating Pedestrian Data in Real Scene and Beyond
• Learning to Disentangle Scenes for Person Re-identification
• Horizontal Pyramid Matching for Person Re-identification
• Improved Person Re-Identification Based on Saliency and Semantic Parsing with Deep Neural Network Models
• Semi-supervised Feature Learning For Improving Writer Identification
• Let Features Decide for Themselves: Feature Mask Network for Person Re-identification
• Diversity Regularized Spatiotemporal Attention for Video-based Person Re-identification
• Self-similarity Grouping: A Simple Unsupervised Cross Domain Adaptation Approach for Person Re-identification
• High-Order Information Matters: Learning Relation and Topology for Occluded Person Re-Identification
• Domain Generalization with MixStyle
• Relation-Aware Global Attention for Person Re-identification
• Domain Adaptive Person Re-Identification via Coupling Optimization
• Learning Shape Representations for Clothing Variations in Person Re-Identification
• Pose-guided Visible Part Matching for Occluded Person ReID
• Batch DropBlock Network for Person Re-identification and Beyond
• Unsupervised Person Re-identification via Softened Similarity Learning
• Style Normalization and Restitution for Generalizable Person Re-identification
• CA3Net: Contextual-Attentional Attribute-Appearance Network for Person Re-Identification
• Progressive DARTS: Bridging the Optimization Gap for NAS in the Wild
• Densely Semantically Aligned Person Re-Identification
• Beyond Human Parts: Dual Part-Aligned Representations for Person Re-Identification
• Frustratingly Easy Person Re-Identification: Generalizing Person Re-ID in Practice
• Progressive Sample Mining and Representation Learning for One-Shot Person Re-identification with Adversarial Samples
• Single-Label Multi-Class Image Classification by Deep Logistic Regression
• Recover and Identify: A Generative Dual Model for Cross-Resolution Person Re-Identification
• Re-Identification with Consistent Attentive Siamese Networks
• Dissecting Person Re-identification from the Viewpoint of Viewpoint
• View Confusion Feature Learning for Person Re-identification
• Multi-Granularity Reference-Aided Attentive Feature Aggregation for Video-based Person Re-identification
• Sequence-based Person Attribute Recognition with Joint CTC-Attention Model
• SBSGAN: Suppression of Inter-Domain Background Shift for Person Re-Identification
• Semantics-Aligned Representation Learning for Person Re-identification
• Semi-Supervised Domain Generalizable Person Re-Identification
• When Person Re-identification Meets Changing Clothes
• A Person Re-identification Data Augmentation Method with Adversarial Defense Effect
• Deep Domain-Adversarial Image Generation for Domain Generalisation
• Unity Style Transfer for Person Re-Identification
• Uncertainty-Aware Multi-Shot Knowledge Distillation for Image-Based Object Re-Identification
• Backbone Can Not be Trained at Once: Rolling Back to Pre-trained Network for Person Re-Identification
• Unsupervised Data Uncertainty Learning in Visual Retrieval Systems
• Exploring Uncertainty in Conditional Multi-Modal Retrieval Systems
• Identity Preserving Generative Adversarial Network for Cross-Domain Person Re-identification
• Deep Active Learning for Video-based Person Re-identification
• In Defense of the Classification Loss for Person Re-Identification
• Faster Person Re-Identification
• Support Neighbor Loss for Person Re-Identification
• Universal Person Re-Identification
• EgoReID Dataset: Person Re-identification in Videos Acquired by Mobile Devices with First-Person Point-of-View
• Deep Person Re-Identification with Improved Embedding and Efficient Training
• Triplet Online Instance Matching Loss for Person Re-identification
• Towards Precise Intra-camera Supervised Person Re-identification
• Weakly supervised discriminative feature learning with state information for person identification
• Person Re-identification with Bias-controlled Adversarial Training
• Enhancing Person Re-identification in a Self-trained Subspace
• Semantic Consistency and Identity Mapping Multi-Component Generative Adversarial Network for Person Re-Identification
• Cross-Resolution Person Re-identification with Deep Antithetical Learning
• Pseudo-positive regularization for deep person re-identification
• Regularizing Deep Hashing Networks Using GAN Generated Fake Images
• Adversarial Multi-scale Feature Learning for Person Re-identification