Cross-Domain Weakly-Supervised Object Detection Through Progressive Domain Adaptation (2018-03-30T00:00:00.000000Z)

TL;DR

This paper proposes a two-step progressive domain adaptation technique by fine-tuning the detector on two types of artificially and automatically generated samples, and achieves an improvement of approximately 5 to 20 percentage points in terms of mean average precision (mAP) compared to the best-performing baselines.

Abstract

Can we detect common objects in a variety of image domains without instance-level annotations? In this paper, we present a framework for a novel task, cross-domain weakly supervised object detection, which addresses this question. For this paper, we have access to images with instance-level annotations in a source domain (e.g., natural image) and images with image-level annotations in a target domain (e.g., watercolor). In addition, the classes to be detected in the target domain are all or a subset of those in the source domain. Starting from a fully supervised object detector, which is pre-trained on the source domain, we propose a two-step progressive domain adaptation technique by fine-tuning the detector on two types of artificially and automatically generated samples. We test our methods on our newly collected datasets1 containing three image domains, and achieve an improvement of approximately 5 to 20 percentage points in terms of mean average precision (mAP) compared to the best-performing baselines.

Authors

Naoto Inoue

4 papers

Ryosuke Furuta

4 papers

T. Yamasaki

6 papers

TL;DR

Abstract

Authors

References42 items

ChainerCV: a Library for Deep Learning in Computer Vision

Extreme Clicking for Efficient Object Annotation

Focal Loss for Dense Object Detection

BAM! The Behance Artistic Media Dataset for Recognition Beyond Photography

Multiple Instance Detection Network with Online Instance Classifier Refinement

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

Adversarial Discriminative Domain Adaptation

YOLO9000: Better, Faster, Stronger

Learning from Simulated and Unsupervised Images through Adversarial Training

Unsupervised Pixel-Level Domain Adaptation with Generative Adversarial Networks

Detecting People in Artwork with CNNs

ContextLocNet: Context-Aware Deep Network Models for Weakly Supervised Localization

Learning Aligned Cross-Modal Representations from Weakly Aligned Data

Weakly Supervised Object Localization with Progressive Domain Adaptation

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

Unsupervised Domain Adaptation with Residual Transfer Networks

SSD: Single Shot MultiBox Detector

Weakly Supervised Deep Detection Networks

Weakly supervised object detection with convex clustering

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

Domain-Adversarial Training of Neural Networks

Webly Supervised Learning of Convolutional Networks

Fast R-CNN

Learning Transferable Features with Deep Adaptation Networks

Adam: A Method for Stochastic Optimization

Detector discovery in the wild: Joint multiple instance and representation learning

Learning Graphs to Model Visual Objects across Different Depictive Styles

Weakly-supervised Discovery of Visual Pattern Configurations

Multi-fold MIL Training for Weakly Supervised Object Localization

Microsoft COCO: Common Objects in Context

On learning to localize objects with minimal supervision

Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation

Selective Search for Object Recognition

Crowdsourcing Annotations for Visual Object Detection

Diagnosing Error in Object Detectors

A Kernel Two-Sample Test

Object Detection with Discriminatively Trained Part Based Models

Et al

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

Chainer : a Next-Generation Open Source Framework for Deep Learning

International Journal of Computer Vision manuscript No. (will be inserted by the editor) The PASCAL Visual Object Classes (VOC) Challenge

Openimages: A public dataset for large-scale multi-label and multi-class image classiﬁcation

Field of Study

Journal Information

Name

Page