Causal interventional training for image recognition
Deep learning models often fit undesired dataset bias in training. In this paper, we formulate the bias using causal inference, which helps us uncover the ever-elusive causalities among the key factors in training, and thus pursue the desired causal effect without the bias. We start from revisiting...
Saved in:
Main Authors: | , , , , |
---|---|
Format: | text |
Language: | English |
Published: |
Institutional Knowledge at Singapore Management University
2023
|
Subjects: | |
Online Access: | https://ink.library.smu.edu.sg/sis_research/6743 https://ink.library.smu.edu.sg/context/sis_research/article/7746/viewcontent/CIT_final.pdf |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Singapore Management University |
Language: | English |
id |
sg-smu-ink.sis_research-7746 |
---|---|
record_format |
dspace |
spelling |
sg-smu-ink.sis_research-77462023-06-26T08:25:32Z Causal interventional training for image recognition QIN, Wei ZHANG, Hanwang HONG, Richang LIM, Ee-Peng SUN, Qianru Deep learning models often fit undesired dataset bias in training. In this paper, we formulate the bias using causal inference, which helps us uncover the ever-elusive causalities among the key factors in training, and thus pursue the desired causal effect without the bias. We start from revisiting the process of building a visual recognition system, and then propose a structural causal model (SCM) for the key variables involved in dataset collection and recognition model: object, common sense, bias, context, and label prediction. Based on the SCM, one can observe that there are “good” and “bad” biases. Intuitively, in the image where a car is driving on a high way in a desert, the “good” bias denoting the common-sense context is the highway, and the “bad” bias accounting for the noisy context factor is the desert. We tackle this problem with a novel causal interventional training (CIT) approach, where we control the observed context in each object class. We offer theoretical justifications for CIT and validate it with extensive classification experiments on CIFAR-10, CIFAR-100 and ImageNet, e.g., surpassing the standard deep neural networks ResNet-34 and ResNet-50, respectively, by 0.95% and 0.70% accuracies on the ImageNet. Our code is open-sourced on the GitHub https://github.com/qinwei-hfut/CIT. 2023-01-01T08:00:00Z text application/pdf https://ink.library.smu.edu.sg/sis_research/6743 info:doi/10.1109/TMM.2021.3136717 https://ink.library.smu.edu.sg/context/sis_research/article/7746/viewcontent/CIT_final.pdf http://creativecommons.org/licenses/by-nc-nd/4.0/ Research Collection School Of Computing and Information Systems eng Institutional Knowledge at Singapore Management University Image recognition causality causal intervention deep learning ImageNet Databases and Information Systems Graphics and Human Computer Interfaces |
institution |
Singapore Management University |
building |
SMU Libraries |
continent |
Asia |
country |
Singapore Singapore |
content_provider |
SMU Libraries |
collection |
InK@SMU |
language |
English |
topic |
Image recognition causality causal intervention deep learning ImageNet Databases and Information Systems Graphics and Human Computer Interfaces |
spellingShingle |
Image recognition causality causal intervention deep learning ImageNet Databases and Information Systems Graphics and Human Computer Interfaces QIN, Wei ZHANG, Hanwang HONG, Richang LIM, Ee-Peng SUN, Qianru Causal interventional training for image recognition |
description |
Deep learning models often fit undesired dataset bias in training. In this paper, we formulate the bias using causal inference, which helps us uncover the ever-elusive causalities among the key factors in training, and thus pursue the desired causal effect without the bias. We start from revisiting the process of building a visual recognition system, and then propose a structural causal model (SCM) for the key variables involved in dataset collection and recognition model: object, common sense, bias, context, and label prediction. Based on the SCM, one can observe that there are “good” and “bad” biases. Intuitively, in the image where a car is driving on a high way in a desert, the “good” bias denoting the common-sense context is the highway, and the “bad” bias accounting for the noisy context factor is the desert. We tackle this problem with a novel causal interventional training (CIT) approach, where we control the observed context in each object class. We offer theoretical justifications for CIT and validate it with extensive classification experiments on CIFAR-10, CIFAR-100 and ImageNet, e.g., surpassing the standard deep neural networks ResNet-34 and ResNet-50, respectively, by 0.95% and 0.70% accuracies on the ImageNet. Our code is open-sourced on the GitHub https://github.com/qinwei-hfut/CIT. |
format |
text |
author |
QIN, Wei ZHANG, Hanwang HONG, Richang LIM, Ee-Peng SUN, Qianru |
author_facet |
QIN, Wei ZHANG, Hanwang HONG, Richang LIM, Ee-Peng SUN, Qianru |
author_sort |
QIN, Wei |
title |
Causal interventional training for image recognition |
title_short |
Causal interventional training for image recognition |
title_full |
Causal interventional training for image recognition |
title_fullStr |
Causal interventional training for image recognition |
title_full_unstemmed |
Causal interventional training for image recognition |
title_sort |
causal interventional training for image recognition |
publisher |
Institutional Knowledge at Singapore Management University |
publishDate |
2023 |
url |
https://ink.library.smu.edu.sg/sis_research/6743 https://ink.library.smu.edu.sg/context/sis_research/article/7746/viewcontent/CIT_final.pdf |
_version_ |
1770576570323304448 |