ALCN: Adaptive Local Contrast Normalization for Robust Object Detection and 3D Pose Estimation

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

Abstract

To be robust to illumination changes when detecting objects in images, the current
trend is to train a Deep Network with training images captured under many different
lighting conditions. Unfortunately, creating such a training set is very cumbersome, or
sometimes even impossible, for some applications such as 3D pose estimation of specific
objects, which is the application we focus on in this paper. We therefore propose a
novel illumination normalization method that lets us learn to detect objects and estimate
their 3D pose under challenging illumination conditions from very few training samples.
Our key insight is that normalization parameters should adapt to the input image. In
particular, we realized this via a Convolutional Neural Network trained to predict the
parameters of a generalization of the Difference-of-Gaussians method. We show that our
method significantly outperforms standard normalization methods and demonstrate it on
two challenging 3D detection and pose estimation problems.
Original languageEnglish
Title of host publicationBritish Machine Vision Conference
Publication statusPublished - 2017
EventBritish Machine Vision Conference - London, United Kingdom
Duration: 4 Sep 20177 Apr 2018

Conference

ConferenceBritish Machine Vision Conference
CountryUnited Kingdom
CityLondon
Period4/09/177/04/18

Fingerprint

Lighting
Neural networks
Object detection

Cite this

ALCN: Adaptive Local Contrast Normalization for Robust Object Detection and 3D Pose Estimation. / Rad, Mahdi; Roth, Peter M.; Lepetit, Vincent.

British Machine Vision Conference. 2017.

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

Rad, M, Roth, PM & Lepetit, V 2017, ALCN: Adaptive Local Contrast Normalization for Robust Object Detection and 3D Pose Estimation. in British Machine Vision Conference. British Machine Vision Conference, London, United Kingdom, 4/09/17.
@inproceedings{5672dc9a6cab46369e51842c296b9551,
title = "ALCN: Adaptive Local Contrast Normalization for Robust Object Detection and 3D Pose Estimation",
abstract = "To be robust to illumination changes when detecting objects in images, the currenttrend is to train a Deep Network with training images captured under many differentlighting conditions. Unfortunately, creating such a training set is very cumbersome, orsometimes even impossible, for some applications such as 3D pose estimation of specificobjects, which is the application we focus on in this paper. We therefore propose anovel illumination normalization method that lets us learn to detect objects and estimatetheir 3D pose under challenging illumination conditions from very few training samples.Our key insight is that normalization parameters should adapt to the input image. Inparticular, we realized this via a Convolutional Neural Network trained to predict theparameters of a generalization of the Difference-of-Gaussians method. We show that ourmethod significantly outperforms standard normalization methods and demonstrate it ontwo challenging 3D detection and pose estimation problems.",
author = "Mahdi Rad and Roth, {Peter M.} and Vincent Lepetit",
year = "2017",
language = "English",
booktitle = "British Machine Vision Conference",

}

TY - GEN

T1 - ALCN: Adaptive Local Contrast Normalization for Robust Object Detection and 3D Pose Estimation

AU - Rad, Mahdi

AU - Roth, Peter M.

AU - Lepetit, Vincent

PY - 2017

Y1 - 2017

N2 - To be robust to illumination changes when detecting objects in images, the currenttrend is to train a Deep Network with training images captured under many differentlighting conditions. Unfortunately, creating such a training set is very cumbersome, orsometimes even impossible, for some applications such as 3D pose estimation of specificobjects, which is the application we focus on in this paper. We therefore propose anovel illumination normalization method that lets us learn to detect objects and estimatetheir 3D pose under challenging illumination conditions from very few training samples.Our key insight is that normalization parameters should adapt to the input image. Inparticular, we realized this via a Convolutional Neural Network trained to predict theparameters of a generalization of the Difference-of-Gaussians method. We show that ourmethod significantly outperforms standard normalization methods and demonstrate it ontwo challenging 3D detection and pose estimation problems.

AB - To be robust to illumination changes when detecting objects in images, the currenttrend is to train a Deep Network with training images captured under many differentlighting conditions. Unfortunately, creating such a training set is very cumbersome, orsometimes even impossible, for some applications such as 3D pose estimation of specificobjects, which is the application we focus on in this paper. We therefore propose anovel illumination normalization method that lets us learn to detect objects and estimatetheir 3D pose under challenging illumination conditions from very few training samples.Our key insight is that normalization parameters should adapt to the input image. Inparticular, we realized this via a Convolutional Neural Network trained to predict theparameters of a generalization of the Difference-of-Gaussians method. We show that ourmethod significantly outperforms standard normalization methods and demonstrate it ontwo challenging 3D detection and pose estimation problems.

M3 - Conference contribution

BT - British Machine Vision Conference

ER -