Attention Guided Food Recognition via Multi-Stage Local Feature Fusion DOI Open Access

Gonghui Deng,

Dunzhi Wu,

Weizhen Chen

и другие.

Computers, materials & continua/Computers, materials & continua (Print), Год журнала: 2024, Номер 80(2), С. 1985 - 2003

Опубликована: Янв. 1, 2024

The task of food image recognition, a nuanced subset fine-grained grapples with substantial intra-class variation and minimal inter-class differences. These challenges are compounded by the irregular multi-scale nature images. Addressing these complexities, our study introduces an advanced model that leverages multiple attention mechanisms multi-stage local fusion, grounded in ConvNeXt architecture. Our employs hybrid (HA) to pinpoint critical discriminative regions within images, substantially mitigating influence background noise. Furthermore, it fusion (MSLF) module, fostering long-distance dependencies between feature maps at varying stages. This approach facilitates assimilation complementary features across scales, significantly bolstering model's capacity for extraction. we constructed dataset named Roushi60, which consists 60 different categories common meat dishes. Empirical evaluation ETH Food-101, ChineseFoodNet, Roushi60 datasets reveals achieves recognition accuracies 91.12%, 82.86%, 92.50%, respectively. figures not only mark improvement 1.04%, 3.42%, 1.36% over foundational network but also surpass performance most contemporary methods. Such advancements underscore efficacy proposed navigating intricate landscape setting new benchmark field.

Язык: Английский

Image‐based food groups and portion prediction by using deep learning DOI Creative Commons
H. Selçuk Noğay, Nalan Hakime Noğay, Hojjat Adeli

и другие.

Journal of Food Science, Год журнала: 2025, Номер 90(3)

Опубликована: Март 1, 2025

Chronic diseases such as obesity and hypertension due to malnutrition can be prevented by following the appropriate diet, correct diet intake with measuring portion size, developing healthy eating habits. Having a system that automatically measure food consumption is important determine whether individual nutritional needs are being met in order accurately diagnose solve problems, act quickly, minimize risk of cross-cultural diversity foods. In this study, deep learning has been developed implemented for grouping classifying Dishes from Turkish cuisine were chosen sample application testing. The method used convolutional neural network (CNN) models based on image recognition. This study using CNNs classify groups estimate sizes dishes, achieving accuracy rates up 80% group classification 80.47% estimation inclusion data augmentation.

Язык: Английский

Процитировано

0
Multi-layer adaptive spatial-temporal feature fusion network for efficient food image recognition DOI

Sirawan Phiphitphatphaisit,

Olarik Surinta

Expert Systems with Applications, Год журнала: 2024, Номер 255, С. 124834 - 124834

Опубликована: Июль 24, 2024

Язык: Английский

Процитировано

3
Attention Guided Food Recognition via Multi-Stage Local Feature Fusion DOI Open Access

Gonghui Deng,

Dunzhi Wu,

Weizhen Chen

и другие.

Computers, materials & continua/Computers, materials & continua (Print), Год журнала: 2024, Номер 80(2), С. 1985 - 2003

Опубликована: Янв. 1, 2024

The task of food image recognition, a nuanced subset fine-grained grapples with substantial intra-class variation and minimal inter-class differences. These challenges are compounded by the irregular multi-scale nature images. Addressing these complexities, our study introduces an advanced model that leverages multiple attention mechanisms multi-stage local fusion, grounded in ConvNeXt architecture. Our employs hybrid (HA) to pinpoint critical discriminative regions within images, substantially mitigating influence background noise. Furthermore, it fusion (MSLF) module, fostering long-distance dependencies between feature maps at varying stages. This approach facilitates assimilation complementary features across scales, significantly bolstering model's capacity for extraction. we constructed dataset named Roushi60, which consists 60 different categories common meat dishes. Empirical evaluation ETH Food-101, ChineseFoodNet, Roushi60 datasets reveals achieves recognition accuracies 91.12%, 82.86%, 92.50%, respectively. figures not only mark improvement 1.04%, 3.42%, 1.36% over foundational network but also surpass performance most contemporary methods. Such advancements underscore efficacy proposed navigating intricate landscape setting new benchmark field.

Язык: Английский

Процитировано

1