Domain Adaptive Faster R-CNN for Object Detection in the Wild (2018-03-08T00:00:00.000000Z)

TL;DR

This work builds on the recent state-of-the-art Faster R-CNN model, and design two domain adaptation components, on image level and instance level, to reduce the domain discrepancy, based on $$-divergence theory.

Abstract

Object detection typically assumes that training and test data are drawn from an identical distribution, which, however, does not always hold in practice. Such a distribution mismatch will lead to a significant performance drop. In this work, we aim to improve the cross-domain robustness of object detection. We tackle the domain shift on two levels: 1) the image-level shift, such as image style, illumination, etc., and 2) the instance-level shift, such as object appearance, size, etc. We build our approach based on the recent state-of-the-art Faster R-CNN model, and design two domain adaptation components, on image level and instance level, to reduce the domain discrepancy. The two domain adaptation components are based on $$-divergence theory, and are implemented by learning a domain classifier in adversarial training manner. The domain classifiers on different levels are further reinforced with a consistency regularization to learn a domain-invariant region proposal network (RPN) in the Faster R-CNN model. We evaluate our newly proposed approach using multiple datasets including Cityscapes, KITTI, SIM10K, etc. The results demonstrate the effectiveness of our proposed approach for robust object detection in various domain shift scenarios.

Authors

L. Gool

68 papers

Dengxin Dai

12 papers

Wen Li

2 papers

TL;DR

Abstract

Authors

References62 items

Domain Generalization and Adaptation Using Low Rank Exemplar SVMs

ROAD: Reality Oriented Adaptation for Semantic Segmentation of Urban Scenes

Deeper, Broader and Artier Domain Generalization

Open Set Domain Adaptation

Unified Deep Supervised Domain Adaptation and Generalization

Fine-Grained Recognition in the Wild: A Multi-task Domain Adaptation Approach

Semantic Foggy Scene Understanding with Synthetic Data

Associative Domain Adaptation

Curriculum Domain Adaptation for Semantic Segmentation of Urban Scenes

When Unsupervised Domain Adaptation Meets Tensor Representations

Deep Domain Adaptation by Geodesic Distance Minimization

AutoDIAL: Automatic Domain Alignment Layers

DualGAN: Unsupervised Dual Learning for Image-to-Image Translation

Unpaired Image-to-Image Translation Using Cycle-Consistent Adversarial Networks

Learning to Discover Cross-Domain Relations with Generative Adversarial Networks

Unsupervised Image-to-Image Translation Networks

Feature Pyramid Networks for Object Detection

FCNs in the Wild: Pixel-level Adversarial and Constraint-based Adaptation

Driving in the Matrix: Can virtual worlds replace human-generated annotations for real world tasks?

Is Faster R-CNN Doing Well for Pedestrian Detection?

Deep Reconstruction-Classification Networks for Unsupervised Domain Adaptation

Scale-Aware Alignment of Hierarchical Image Segmentation

R-FCN: Object Detection via Region-based Fully Convolutional Networks

The Cityscapes Dataset for Semantic Urban Scene Understanding

Perceptual Losses for Real-Time Style Transfer and Super-Resolution

Instance-Aware Semantic Segmentation via Multi-task Network Cascades

SSD: Single Shot MultiBox Detector

Return of Frustratingly Easy Domain Adaptation

Attention to Scale: Scale-Aware Semantic Image Segmentation

Is image super-resolution helpful for other vision tasks?

Subspace Alignment Based Domain Adaptation for RCNN Detector

Learning scene-specific pedestrian detectors without real data

Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks

Object Detection via a Multi-region and Semantic Segmentation-Aware CNN Model

Fast R-CNN

Learning Transferable Features with Deep Adaptation Networks

Learning Deep Object Detectors from 3D Models

Domain Adaptation of Deformable Part-Based Models

Unsupervised Domain Adaptation by Backpropagation

Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition

Microsoft COCO: Common Objects in Context

OverFeat: Integrated Recognition, Localization and Detection using Convolutional Networks

Unsupervised Visual Domain Adaptation Using Subspace Alignment

Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation

Vision meets robotics: The KITTI dataset

Shifting Weights: Adapting Object Detectors from Image to Video

ImageNet classification with deep convolutional neural networks

Diagnosing Error in Object Detectors

Geodesic flow kernel for unsupervised domain adaptation

Domain Transfer Multiple Kernel Learning

Domain adaptation for object recognition: An unsupervised approach

Unbiased look at dataset bias

What you saw is not what you get: Domain adaptation using asymmetric kernel transforms

Object Detection with Discriminatively Trained Part Based Models

A theory of learning from different domains

Histograms of oriented gradients for human detection

Learning Transferrable Representations for Unsupervised Domain Adaptation

Learning Scene-Specific Pedestrian Detectors without Real Data

From Virtual to Reality: Fast Adaptation of Virtual Object Detectors to Real Domains

Visual event recognition in videos by learning from web data

International Journal of Computer Vision 48(3), 233--254, 2002

Edinburgh Research Explorer The PASCAL Visual Object Classes (VOC) Challenge

Field of Study

Journal Information

Name

Page