Diabetes60 — Inferring Bread Units From Food Images Using Fully Convolutional Neural Networks

Christ, Patrick Ferdinand; Schlecht, Sebastian J.; Ettlinger, Florian; Grün, Felix; Heinle, Christoph; Tatavatry, Sunil; Ahmadi, Seyed-Ahmad; Diepold, Klaus; Menze, Bjoern

doi:10.1109/iccvw.2017.180

Cited by 23 publications

(19 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Evidently, reducing the number of images required to be taken by users would be ergonomically ideal, but this can lead to inaccuracies in quantifying volume due to occlusions that are not visible in a single-view image [108][109][110]120]. To circumvent this issue, researchers have extended the use of depth sensing technology to integrate the aspect of deep learning, allowing 3D depth maps of food objects to be predicted based on visible surfaces [108][109][110]120]. This model works on the assumption that sufficient images and training will allow the system to understand the context of the scene, allowing for camera viewing angles, depth value of food objects, and occluded regions to be extrapolated from the images [108][109][110]120].…”

Section: Depth Mappingmentioning

confidence: 99%

“…To circumvent this issue, researchers have extended the use of depth sensing technology to integrate the aspect of deep learning, allowing 3D depth maps of food objects to be predicted based on visible surfaces [108][109][110]120]. This model works on the assumption that sufficient images and training will allow the system to understand the context of the scene, allowing for camera viewing angles, depth value of food objects, and occluded regions to be extrapolated from the images [108][109][110]120]. Unlike previous applications of deep learning in food volume estimation that relied on relative estimations, this provides an avenue for more absolute volume calculations to be performed.…”

Section: Depth Mappingmentioning

confidence: 99%

“…Christ et al developed a system that worked on similar principles, but instead of deriving absolute food volume, the system was designed to assist diabetics with estimating servings of carbohydrates, referred to in the paper as bread units (BU) [120]. This was done by training the system with a set of RGB-D images that had been annotated with the corresponding number of contained BUs by human experts.…”

Section: Depth Mappingmentioning

confidence: 99%

“…This may present a degree of noticeable inaccuracy in volume calculations in cases where food is positioned on sloped surfaces such as in a bowl, items that are overlapping each other, or in cases of irregularly shaped foods such as chicken wings. Evidently, reducing the number of images required to be taken by users would be ergonomically ideal, but this can lead to inaccuracies in quantifying volume due to occlusions that are not visible in a single-view image [108][109][110]120]. To circumvent this issue, researchers have extended the use of depth sensing technology to integrate the aspect of deep learning, allowing 3D depth maps of food objects to be predicted based on visible surfaces [108][109][110]120].…”

Section: Depth Mappingmentioning

confidence: 99%

See 3 more Smart Citations

Current Developments in Digital Quantitative Volume Estimation for the Optimisation of Dietary Assessment

et al. 2020

View full text Add to dashboard Cite

Obesity is a global health problem with wide-reaching economic and social implications. Nutrition surveillance systems are essential to understanding and addressing poor dietary practices. However, diets are incredibly diverse across populations and an accurate diagnosis of individualized nutritional issues is challenging. Current tools used in dietary assessment are cumbersome for users, and are only able to provide approximations of dietary information. Given the need for technological innovation, this paper reviews various novel digital methods for food volume estimation and explores the potential for adopting such technology in the Southeast Asian context. We discuss the current approaches to dietary assessment, as well as the potential opportunities that digital health can offer to the field. Recent advances in optics, computer vision and deep learning show promise in advancing the field of quantitative dietary assessment. The ease of access to the internet and the availability of smartphones with integrated cameras have expanded the toolsets available, and there is potential for automated food volume estimation to be developed and integrated as part of a digital dietary assessment tool. Such a tool may enable public health institutions to be able to gather an effective nutritional insight and combat the rising rates of obesity in the region.

show abstract

Section: Depth Mappingmentioning

confidence: 99%

Section: Depth Mappingmentioning

confidence: 99%

Section: Depth Mappingmentioning

confidence: 99%

Section: Depth Mappingmentioning

confidence: 99%

See 2 more Smart Citations

Current Developments in Digital Quantitative Volume Estimation for the Optimisation of Dietary Assessment

et al. 2020

View full text Add to dashboard Cite

show abstract

“…A first attempt has been reported by Allegra et al [26], which applied a SegNet-based [27] CNN architecture for single food image depth prediction. Recently, Christ et al [28] proposed a CNN architecture with skip connections for food image depth prediction.…”

Section: Depth Predictionmentioning

confidence: 99%

A multi-task learning approach for meal assessment

Allegra

Anthimopoulos

et al. 2018

Proceedings of the Joint Workshop on Multimedia for Cooking and Eating Activities and Multimedia Assisted Dietary Management

View full text Add to dashboard Cite

Key role in the prevention of diet-related chronic diseases plays the balanced nutrition together with a proper diet. The conventional dietary assessment methods are time-consuming, expensive and prone to errors. New technology-based methods that provide reliable and convenient dietary assessment, have emerged during the last decade. The advances in the field of computer vision permitted the use of meal image to assess the nutrient content usually through three steps: food segmentation, recognition and volume estimation. In this paper, we propose a use one RGB meal image as input to a multi-task learning based Convolutional Neural Network (CNN). The proposed approach achieved outstanding performance, while a comparison with state-of-the-art methods indicated that the proposed approach exhibits clear advantage in accuracy, along with a massive reduction of processing time.

show abstract

Investigation of Methodologies of Food Volume Estimation and Dataset for Image-Based Dietary Assessment

Kadam

Petkar

Phansalkar

2021

Information and Communication Technology for Competitive Strategies (ICTCS 2020)

View full text Add to dashboard Cite

Diabetes60 — Inferring Bread Units From Food Images Using Fully Convolutional Neural Networks

Cited by 23 publications

References 27 publications

Current Developments in Digital Quantitative Volume Estimation for the Optimisation of Dietary Assessment

Current Developments in Digital Quantitative Volume Estimation for the Optimisation of Dietary Assessment

A multi-task learning approach for meal assessment

Investigation of Methodologies of Food Volume Estimation and Dataset for Image-Based Dietary Assessment

Contact Info

Product

Resources

About