Learning Deep Features for Discriminative Localization (2015-12-14T00:00:00.000000Z)

TL;DR

This work revisits the global average pooling layer proposed in [13], and sheds light on how it explicitly enables the convolutional neural network (CNN) to have remarkable localization ability despite being trained on imagelevel labels.

Abstract

In this work, we revisit the global average pooling layer proposed in [13], and shed light on how it explicitly enables the convolutional neural network (CNN) to have remarkable localization ability despite being trained on imagelevel labels. While this technique was previously proposed as a means for regularizing training, we find that it actually builds a generic localizable deep representation that exposes the implicit attention of CNNs on an image. Despite the apparent simplicity of global average pooling, we are able to achieve 37.1% top-5 error for object localization on ILSVRC 2014 without training on any bounding box annotation. We demonstrate in a variety of experiments that our network is able to localize the discriminative image regions despite just being trained for solving classification task1.

Authors

A. Khosla

10 papers

Bolei Zhou

10 papers

Àgata Lapedriza

7 papers

TL;DR

Abstract

Authors

References37 items

Simple Baseline for Visual Question Answering

Inverting Visual Representations with Convolutional Networks

Inverting Convolutional Networks with Convolutional Networks

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

Weakly Supervised Object Localization with Multi-Fold Multiple Instance Learning

Object Detectors Emerge in Deep Scene CNNs

Learning Deep Features for Scene Recognition using Places Database

Understanding deep image representations by inverting them

From image-level to pixel-level labeling with Convolutional Networks

ConceptLearner: Discovering visual concepts from weakly labeled image collections

Going deeper with convolutions

Self-taught object localization with deep networks

Very Deep Convolutional Networks for Large-Scale Image Recognition

ImageNet Large Scale Visual Recognition Challenge

Local Alignments for Fine-Grained Categorization

Part-Based R-CNNs for Fine-Grained Category Detection

Learning and Transferring Mid-level Image Representations Using Convolutional Neural Networks

CNN Features Off-the-Shelf: An Astounding Baseline for Recognition

OverFeat: Integrated Recognition, Localization and Detection using Convolutional Networks

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

Network In Network

Deformable Part Descriptors for Fine-Grained Recognition and Attribute Prediction

Visualizing and Understanding Convolutional Networks

Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation

DeCAF: A Deep Convolutional Activation Feature for Generic Visual Recognition

ImageNet classification with deep convolutional neural networks

SUN attribute database: Discovering, annotating, and recognizing scene attributes

Human action recognition by learning bases of action attributes and parts

End-to-end scene text recognition

Caltech-UCSD Birds 200

SUN database: Large-scale scene recognition from abbey to zoo

Recognizing indoor scenes

LIBLINEAR: A Library for Large Linear Classification

What, where and who? Classifying events by scene and object recognition

Caltech-256 Object Category Dataset

Beyond Bags of Features: Spatial Pyramid Matching for Recognizing Natural Scene Categories

Learning Generative Visual Models from Few Training Examples: An Incremental Bayesian Approach Tested on 101 Object Categories

Field of Study

Journal Information

Name

Page

Venue Information

Name

Type

URL

Alternate Names