Long-Tailed Food Classification
Food classification serves as the basic step of image-based dietary assessment to predict the types of foods in each input image. However, foods in real-world scenarios are typically long-tail distributed, where a small number of food types are consumed more frequently than others, which causes a se...
Main Authors: | , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2023-06-01
|
Series: | Nutrients |
Subjects: | |
Online Access: | https://www.mdpi.com/2072-6643/15/12/2751 |
_version_ | 1797593205940682752 |
---|---|
author | Jiangpeng He Luotao Lin Heather A. Eicher-Miller Fengqing Zhu |
author_facet | Jiangpeng He Luotao Lin Heather A. Eicher-Miller Fengqing Zhu |
author_sort | Jiangpeng He |
collection | DOAJ |
description | Food classification serves as the basic step of image-based dietary assessment to predict the types of foods in each input image. However, foods in real-world scenarios are typically long-tail distributed, where a small number of food types are consumed more frequently than others, which causes a severe class imbalance issue and hinders the overall performance. In addition, none of the existing long-tailed classification methods focus on food data, which can be more challenging due to the inter-class similarity and intra-class diversity between food images. In this work, two new benchmark datasets for long-tailed food classification are introduced, including Food101-LT and VFN-LT, where the number of samples in VFN-LT exhibits real-world long-tailed food distribution. Then, a novel two-phase framework is proposed to address the problem of class imbalance by (1) undersampling the head classes to remove redundant samples along with maintaining the learned information through knowledge distillation and (2) oversampling the tail classes by performing visually aware data augmentation. By comparing our method with existing state-of-the-art long-tailed classification methods, we show the effectiveness of the proposed framework, which obtains the best performance on both Food101-LT and VFN-LT datasets. The results demonstrate the potential to apply the proposed method to related real-life applications. |
first_indexed | 2024-03-11T02:04:39Z |
format | Article |
id | doaj.art-c9e5579b52114a888093ff08c97c9732 |
institution | Directory Open Access Journal |
issn | 2072-6643 |
language | English |
last_indexed | 2024-03-11T02:04:39Z |
publishDate | 2023-06-01 |
publisher | MDPI AG |
record_format | Article |
series | Nutrients |
spelling | doaj.art-c9e5579b52114a888093ff08c97c97322023-11-18T11:57:05ZengMDPI AGNutrients2072-66432023-06-011512275110.3390/nu15122751Long-Tailed Food ClassificationJiangpeng He0Luotao Lin1Heather A. Eicher-Miller2Fengqing Zhu3Elmore Family School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN 47907, USADepartment of Nutrition Science, Purdue University, West Lafayette, IN 47907, USADepartment of Nutrition Science, Purdue University, West Lafayette, IN 47907, USAElmore Family School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN 47907, USAFood classification serves as the basic step of image-based dietary assessment to predict the types of foods in each input image. However, foods in real-world scenarios are typically long-tail distributed, where a small number of food types are consumed more frequently than others, which causes a severe class imbalance issue and hinders the overall performance. In addition, none of the existing long-tailed classification methods focus on food data, which can be more challenging due to the inter-class similarity and intra-class diversity between food images. In this work, two new benchmark datasets for long-tailed food classification are introduced, including Food101-LT and VFN-LT, where the number of samples in VFN-LT exhibits real-world long-tailed food distribution. Then, a novel two-phase framework is proposed to address the problem of class imbalance by (1) undersampling the head classes to remove redundant samples along with maintaining the learned information through knowledge distillation and (2) oversampling the tail classes by performing visually aware data augmentation. By comparing our method with existing state-of-the-art long-tailed classification methods, we show the effectiveness of the proposed framework, which obtains the best performance on both Food101-LT and VFN-LT datasets. The results demonstrate the potential to apply the proposed method to related real-life applications.https://www.mdpi.com/2072-6643/15/12/2751food classificationlong-tail distributionimage-based dietary assessmentbenchmark datasetsfood consumption frequencyneural networks |
spellingShingle | Jiangpeng He Luotao Lin Heather A. Eicher-Miller Fengqing Zhu Long-Tailed Food Classification Nutrients food classification long-tail distribution image-based dietary assessment benchmark datasets food consumption frequency neural networks |
title | Long-Tailed Food Classification |
title_full | Long-Tailed Food Classification |
title_fullStr | Long-Tailed Food Classification |
title_full_unstemmed | Long-Tailed Food Classification |
title_short | Long-Tailed Food Classification |
title_sort | long tailed food classification |
topic | food classification long-tail distribution image-based dietary assessment benchmark datasets food consumption frequency neural networks |
url | https://www.mdpi.com/2072-6643/15/12/2751 |
work_keys_str_mv | AT jiangpenghe longtailedfoodclassification AT luotaolin longtailedfoodclassification AT heatheraeichermiller longtailedfoodclassification AT fengqingzhu longtailedfoodclassification |