Perceptual Generative Adversarial Networks for Small Object Detection

Abstract

Detecting small objects is notoriously challenging due to their low resolution and noisy representation. Existing object detection pipelines usually detect small objects through learning representations of all the objects at multiple scales. However, the performance gain of such ad hoc architectures is usually limited to pay off the computational cost. In this work, we address the small object detection problem by developing a single architecture that internally lifts representations of small objects to "super-resolved" ones, achieving similar characteristics as large objects and thus more discriminative for detection. For this purpose, we propose a new Perceptual Generative Adversarial Network (Perceptual GAN) model that improves small object detection through narrowing representation difference of small objects from the large ones. Specifically, its generator learns to transfer perceived poor representations of the small objects to super-resolved ones that are similar enough to real large objects to fool a competing discriminator. Meanwhile its discriminator competes with the generator to identify the generated representation and imposes an additional perceptual requirement - generated representations of small objects must be beneficial for detection purpose - on the generator. Extensive evaluations on the challenging Tsinghua-Tencent 100K and the Caltech benchmark well demonstrate the superiority of Perceptual GAN in detecting small objects, including traffic signs and pedestrians, over well-established state-of-the-arts.

References in corpus (15)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
• Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
• Caffe: Convolutional Architecture for Fast Feature Embedding
• Deep Generative Image Models using a Laplacian Pyramid of Adversarial Networks
• Rich feature hierarchies for accurate object detection and semantic segmentation
• Ten Years of Pedestrian Detection, What Have We Learned?
• How Far are We from Solving Pedestrian Detection?
• Combining Markov Random Fields and Convolutional Neural Networks for Image Synthesis
• Inside-Outside Net: Detecting Objects in Context with Skip Pooling and Recurrent Neural Networks
• Filtered Channel Features for Pedestrian Detection
• Learning Complexity-Aware Cascades for Deep Pedestrian Detection
• Pedestrian Detection with Unsupervised Multi-Stage Feature Learning
• A novel pLSA based Traffic Signs Classification System
• Pedestrian Detection aided by Deep Learning Semantic Tasks

Cited by in corpus (21)

• How Generative Adversarial Networks and Their Variants Work: An Overview
• Deep Learning for UAV-based Object Detection and Tracking: A Survey
• A Review on Generative Adversarial Networks: Algorithms, Theory, and Applications
• Augmentation for small object detection
• CM-GANs: Cross-modal Generative Adversarial Networks for Common Representation Learning
• Decoupled Classification Refinement: Hard False Positive Suppression for Object Detection
• Learning to Generate Time-Lapse Videos Using Multi-Stage Dynamic Generative Adversarial Networks
• Toward Characteristic-Preserving Image-based Virtual Try-On Network
• Adversarial Discriminative Heterogeneous Face Recognition
• A Survey on Deep Domain Adaptation and Tiny Object Detection Challenges, Techniques and Datasets
• GestureGAN for Hand Gesture-to-Gesture Translation in the Wild
• Delving into Robust Object Detection from Unmanned Aerial Vehicles: A Deep Nuisance Disentanglement Approach
• SCH-GAN: Semi-supervised Cross-modal Hashing by Generative Adversarial Network
• Adversarial Occlusion-aware Face Detection
• Density Map Guided Object Detection in Aerial Images
• Ebola Optimization Search Algorithm (EOSA): A new metaheuristic algorithm based on the propagation model of Ebola virus disease
• Segmentation Guided Image-to-Image Translation with Adversarial Networks
• Detecting Heads using Feature Refine Net and Cascaded Multi-Scale Architecture
• Adaptive Temporal Encoding Network for Video Instance-level Human Parsing
• LoGAN: Generating Logos with a Generative Adversarial Neural Network Conditioned on color
• ProbaNet: Proposal-balanced Network for Object Detection