Hide-and-Seek: Forcing a Network to be Meticulous for Weakly-Supervised Object and Action Localization (2017-04-13T00:00:00.000000Z)

TL;DR

The key idea is to hide patches in a training image randomly, forcing the network to seek other relevant parts when the most discriminative part is hidden, which obtains superior performance compared to previous methods for weakly-supervised object localization on the ILSVRC dataset.

Abstract

We propose ‘Hide-and-Seek’, a weakly-supervised framework that aims to improve object localization in images and action localization in videos. Most existing weakly-supervised methods localize only the most discriminative parts of an object rather than all relevant parts, which leads to suboptimal performance. Our key idea is to hide patches in a training image randomly, forcing the network to seek other relevant parts when the most discriminative part is hidden. Our approach only needs to modify the input image and can work with any network designed for object localization. During testing, we do not need to hide any patches. Our Hide-and-Seek approach obtains superior performance compared to previous methods for weakly-supervised object localization on the ILSVRC dataset. We also demonstrate that our framework can be easily extended to weakly-supervised action localization.

Authors

Yong Jae Lee

9 papers

Krishna Kumar Singh

1 papers

TL;DR

Abstract

Authors

References63 items

A-Fast-RCNN: Hard Positive Generation via Adversary for Object Detection

Object Region Mining with Adversarial Erasing: A Simple Classification to Semantic Segmentation Approach

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

End-to-End Localization and Ranking for Relative Attributes

Connectionist Temporal Modeling for Weakly Supervised Action Labeling

Context Encoders: Feature Learning by Inpainting

Walk and Learn: Facial Attribute Representation Learning from Egocentric Video and Contextual Data

Track and Transfer: Watching Videos to Simulate Strong Human Supervision for Weakly-Supervised Object Detection

Temporal Action Localization in Untrimmed Videos via Multi-stage CNNs

Learning Deep Features for Discriminative Localization

SSD: Single Shot MultiBox Detector

Discovering the Spatial Extent of Relative Attributes

Weakly Supervised Object Boundaries

End-to-End Learning of Action Detection from Frame Glimpses in Videos

Weakly Supervised Deep Detection Networks

Constrained Convolutional Neural Networks for Weakly Supervised Segmentation

Is object localization for free? - Weakly-supervised learning with convolutional neural networks

DevNet: A Deep Event Network for multimedia event detection and evidence recounting

Spatial Transformer Networks

Fast R-CNN

Temporal Localization of Fine-Grained Actions in Videos by Domain Transfer from Web Images

Weakly Supervised Object Localization with Multi-Fold Multiple Instance Learning

Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift

Convolutional feature masking for joint object and stuff segmentation

Learning Spatiotemporal Features with 3D Convolutional Networks

Efficient object localization using Convolutional Networks

Fully convolutional networks for semantic segmentation

Going deeper with convolutions

Self-taught object localization with deep networks

Weakly Supervised Object Localization with Latent Category Learning

Hipster Wars: Discovering Elements of Fashion Styles

ImageNet Large Scale Visual Recognition Challenge

Simultaneous Detection and Segmentation

Weakly Supervised Action Labeling in Videos under Ordering Constraints

Weakly-supervised Discovery of Visual Pattern Configurations

Multi-fold MIL Training for Weakly Supervised Object Localization

Large-Scale Video Classification with Convolutional Neural Networks

Efficient Feature Extraction, Encoding, and Classification for Action Recognition

On learning to localize objects with minimal supervision

Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps

Towards Understanding Action Recognition

PANDA: Pose Aligned Networks for Deep Attribute Modeling

Visualizing and Understanding Convolutional Networks

Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation

Weakly Supervised Learning for Attribute Localization in Outdoor Scenes

ImageNet classification with deep convolutional neural networks

In Defence of Negative Mining for Annotating Weakly Labelled Data

Learning object class detectors from weakly annotated video

Efficient activity detection with max-subgraph search

Discovering localized attributes for fine-grained recognition

Automatic Attribute Discovery and Characterization from Noisy Web Data

Automatic annotation of human actions in video

Learning realistic human actions from movies

Weakly Supervised Learning of Part-Based Spatial Models for Visual Object Recognition

Object class recognition by unsupervised scale-invariant learning

Unsupervised Learning of Models for Recognition

Dropout: a simple way to prevent neural networks from overfitting

Weakly Supervised Object Detection with Posterior Regularization

THUMOS challenge: Action recognition with a large number of classes

Reg-ularization of neural network using dropconnect

F is partially within a hidden patch (Fig. 3, green box)

Author manuscript, published in "International Conference on Computer Vision (2013)" Action Recognition with Improved Trajectories

Efﬁcient feature extraction, encoding and classiﬁcation for action recognition

Field of Study

Journal Information

Name

Page

Venue Information

Name

Type

URL

Alternate Names