Pedestrian-Synthesis-GAN: Generating Pedestrian Data in Real Scene and Beyond

Abstract

State-of-the-art pedestrian detection models have achieved great success in many benchmarks. However, these models require lots of annotation information and the labeling process usually takes much time and efforts. In this paper, we propose a method to generate labeled pedestrian data and adapt them to support the training of pedestrian detectors. The proposed framework is built on the Generative Adversarial Network (GAN) with multiple discriminators, trying to synthesize realistic pedestrians and learn the background context simultaneously. To handle the pedestrians of different sizes, we adopt the Spatial Pyramid Pooling (SPP) layer in the discriminator. We conduct experiments on two benchmarks. The results show that our framework can smoothly synthesize pedestrians on background images of variations and different levels of details. To quantitatively evaluate our approach, we add the generated samples into training data of the baseline pedestrian detectors and show the synthetic images are able to improve the detectors' performance.

arXiv commentv2.0,adding supplementary

References in corpus (14)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Conditional Generative Adversarial Nets
• Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
• Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition
• Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks
• Improved Training of Wasserstein GANs
• Pose Guided Person Image Generation
• YOLO9000: Better, Faster, Stronger
• Unsupervised Cross-Domain Image Generation
• MMD GAN: Towards Deeper Understanding of Moment Matching Network
• Unlabeled Samples Generated by GAN Improve the Person Re-identification Baseline in vitro
• Semi-Supervised Learning with Context-Conditional Generative Adversarial Networks
• How Far are We from Solving Pedestrian Detection?
• Generative Adversarial Networks as Variational Training of Energy Based Models

Cited by in corpus (19)

• TransGAN: Two Pure Transformers Can Make One Strong GAN, and That Can Scale Up
• SESAME: Semantic Editing of Scenes by Adding, Manipulating or Erasing Objects
• Deep Person Generation: A Survey from the Perspective of Face, Pose and Cloth Synthesis
• Feature Calibration Network for Occluded Pedestrian Detection
• CERL: A Unified Optimization Framework for Light Enhancement with Realistic Noise
• Learning More with Less: GAN-based Medical Image Augmentation
• Human Image Generation: A Comprehensive Survey
• Synthesizing Long-Term 3D Human Motion and Interaction in 3D Scenes
• BachGAN: High-Resolution Image Synthesis from Salient Object Layout
• PMC-GANs: Generating Multi-Scale High-Quality Pedestrian with Multimodal Cascaded GANs
• Putting Humans in a Scene: Learning Affordance in 3D Indoor Environments
• Generating 3D People in Scenes without People
• Bayesian Cycle-Consistent Generative Adversarial Networks via Marginalizing Latent Sampling
• Learning More with Less: Conditional PGGAN-based Data Augmentation for Brain Metastases Detection Using Highly-Rough Annotation on MR Images
• Inserting Videos into Videos
• Multimodal Image Outpainting With Regularized Normalized Diversification
• Hidden Footprints: Learning Contextual Walkability from 3D Human Trails
• Pathology-Aware Generative Adversarial Networks for Medical Image Augmentation
• Attention-based Adversarial Appearance Learning of Augmented Pedestrians