Food Portion Size Estimation Research Articles

Food Portion Size Estimation for Developing a Healthy Eating Index: Influence of Unit Count and Energy Density

Estimating food consumption is often influenced by the intricate cognitive processes involved in determining portion sizes. Understanding the factors influencing these processes can enhance the accuracy of dietary assessments. This study aims to evaluate the ability of participants to estimate portion sizes of common foods and to identify estimation errors. Methods: This study used a cross-sectional approach involving 35 participants, who estimated the portions of various foods, measured in grams or milliliters, based on memory. The magnitude and direction of estimation errors were analyzed. Underestimation was defined as estimated quantities (EQ) less than 90% of the actual weighed quantity (WQ), while overestimation was defined as EQ greater than 110% of the WQ. Results: Only 15.59% of the estimates were considered accurate (approximately 10% of the actual weight). Animal proteins, vegetables, and sugars were the most frequently underestimated, while staple foods, vegetable proteins, fruits, vegetables, and fats/oils were the most commonly overestimated. Conclusion: This study revealed a low proportion of accurate food portion size estimates, with a tendency to overestimate low-energy foods and underestimate high-energy foods.

Development and evaluation of food photograph series software for portion size estimation among urban North Indian adults

BACKGROUND: Accurate population specific aids are needed for food portion size estimation in diet assessment. OBJECTIVE: To develop and evaluate a food photograph series software for accurate portion size estimation among urban North Indian adults. METHODS: A preliminary food consumption survey was conducted on 60 adults (25–60 years, male = female = 30) selected purposively from urban Delhi. A range of portion sizes of commonly consumed food items were standardized and photographed at 45 degree angle. An online portion size estimation tool was developed comprising data on 368 recipes and photographs for 313 food items. Real time (part A) and recalled estimation after 24 hours (part B) was done for pre-weighed food portions by participants (n = 30, 25–60 years, male = female = 15) using the software. RESULTS: The % mean relative error in estimation of 24 pre-weighed food portions was –5±22.61% and 13.07±53.22% for parts A and B respectively. Significant correlation between actual and estimated food portions was observed (r > 0.90, p < 0.0001). Correct picture was selected by participants in≥60% estimations while correct or adjacent picture was selected in≥89% estimations. CONCLUSIONS: The software is a useful aid for dietitians and other nutrition professionals for accurate portion size estimation among urban North Indian adults.

Estimation of food portion sizes in women of childbearing age and young children in Ouagadougou (Burkina Faso) using a food photography atlas and salted replicas: Comparison with weighed records.

Quantitative assessment of foods consumed when using 24-hour dietary recall requires accurate tools to estimate portion sizes. Therefore, we developed a food portion photography atlas with age-appropriate portion sizes for 11 foods frequently consumed by young children (sizes for 6-11-month- and for 12-23-month-old children) and women of childbearing age in Ouagadougou, Burkina Faso capital. We then compared the accuracy and precision of portion estimation with this atlas and with salted replicas relative to weighed records (the reference). After weighing, we randomly assigned food portions to 67 women and their children. The next day, women estimated the served portions and leftovers by recall using the atlas and then salted replicas (n = 1156 measurements, ranging from 19 to 113 for each food). For most food types, the portions estimated with the atlas and salted replicas were positively correlated and showed good concordance with the weighed records. However, accuracy and precision varied in function of the estimation method, food type, and age group. The mean crude differences ranged from -28 to +12g (with errors in absolute values from 24 to 69%) for children, and from -32 to +44g (errors from 17 to 56%) for women. The atlas-based method showed the lowest Lin's concordances (coefficients of 0.1 to 0.2) for the leafy vegetable dish, meat, and fish in 12-23-month-old children. Bland-Altman plots indicated that the salted replicas allowed estimating the consumed portions with fewer errors than the photographic atlas (56 to 91% vs 46 to 79% between the limits of ±50%). Our study highlights that mothers have difficulties in perceiving the quantities of food consumed by their children. Our findings also indicate that the food atlas could be used in food consumption surveys when salted replicas are not available for all food types.

Estimation of Plate and Bowl Dimensions for Food Portion Size Assessment in a Wearable Sensor System.

Automatic food portion size estimation (FPSE) with minimal user burden is a challenging task. Most of the existing FPSE methods use fiducial markers and/or virtual models as dimensional references. An alternative approach is to estimate the dimensions of the eating containers prior to estimating the portion size. In this article, we propose a wearable sensor system (the automatic ingestion monitor integrated with a ranging sensor) and a related method for the estimation of dimensions of plates and bowls. The contributions of this study are: 1) the model eliminates the need for fiducial markers; 2) the camera system [automatic ingestion monitor version 2 (AIM-2)] is not restricted in terms of positioning relative to the food item; 3) our model accounts for radial lens distortion caused due to lens aberrations; 4) a ranging sensor directly gives the distance between the sensor and the eating surface; 5) the model is not restricted to circular plates; and 6) the proposed system implements a passive method that can be used for assessment of container dimensions with minimum user interaction. The error rates (mean ± std. dev) for dimension estimation were 2.01% ± 4.10% for plate widths/diameters, 2.75% ± 38.11% for bowl heights, and 4.58% ± 6.78% for bowl diameters.

Food Image Segmentation Using Multi-Modal Imaging Sensors with Color and Thermal Data

Sensor-based food intake monitoring has become one of the fastest-growing fields in dietary assessment. Researchers are exploring imaging-sensor-based food detection, food recognition, and food portion size estimation. A major problem that is still being tackled in this field is the segmentation of regions of food when multiple food items are present, mainly when similar-looking foods (similar in color and/or texture) are present. Food image segmentation is a relatively under-explored area compared with other fields. This paper proposes a novel approach to food imaging consisting of two imaging sensors: color (Red-Green-Blue) and thermal. Furthermore, we propose a multi-modal four-Dimensional (RGB-T) image segmentation using a k-means clustering algorithm to segment regions of similar-looking food items in multiple combinations of hot, cold, and warm (at room temperature) foods. Six food combinations of two food items each were used to capture RGB and thermal image data. RGB and thermal data were superimposed to form a combined RGB-T image and three sets of data (RGB, thermal, and RGB-T) were tested. A bootstrapped optimization of within-cluster sum of squares (WSS) was employed to determine the optimal number of clusters for each case. The combined RGB-T data achieved better results compared with RGB and thermal data, used individually. The mean ± standard deviation (std. dev.) of the F1 score for RGB-T data was 0.87 ± 0.1 compared with 0.66 ± 0.13 and 0.64 ± 0.39, for RGB and Thermal data, respectively.

FOODCAM: A Novel Structured Light-Stereo Imaging System for Food Portion Size Estimation.

Imaging-based methods of food portion size estimation (FPSE) promise higher accuracies compared to traditional methods. Many FPSE methods require dimensional cues (fiducial markers, finger-references, object-references) in the scene of interest and/or manual human input (wireframes, virtual models). This paper proposes a novel passive, standalone, multispectral, motion-activated, structured light-supplemented, stereo camera for food intake monitoring (FOODCAM) and an associated methodology for FPSE that does not need a dimensional reference given a fixed setup. The proposed device integrated a switchable band (visible/infrared) stereo camera with a structured light emitter. The volume estimation methodology focused on the 3-D reconstruction of food items based on the stereo image pairs captured by the device. The FOODCAM device and the methodology were validated using five food models with complex shapes (banana, brownie, chickpeas, French fries, and popcorn). Results showed that the FOODCAM was able to estimate food portion sizes with an average accuracy of 94.4%, which suggests that the FOODCAM can potentially be used as an instrument in diet and eating behavior studies.

The Estimation of Standard Portion Sizes Using Food Photographic Booklet Among University Students in Turkey

Objective: To examine estimation of standard portion size and, the influence of different food plates on the estimation of food portion size, using photograph booklet in students.
 Methods: The sample consisted of 300 students, aged 18-35 years old from different academic departments. The student’s general characteristics were asked by using a questionnaire, and anthropometric measurements were taken. Students were asked to select the standard portion sizes of each food from a photographic booklet, which was developed by the researchers. Dietary guidelines for Turkey were used for assessing the quantity of standard portion size of foods of a photographic booklet (1).
 Results: The findings of the present study indicate that the food group which the most students provided the accurate estimation was protein foods (36.0%), pursued by grains (35.1%), dairy (29.2%), fruits and vegetables group (28.9%). The most and the least accurate estimated foods were found boiled potato (64%) and cheese (3.7%), respectively. The percentages of the accurate estimation of standard portion size for five foods (meatball, meat cubes, shredded chicken, egg, yogurt) in 24 foods were higher in females than men (p

Assessment of bias and associated factors for food portion quantification with photos in Brazil

Food portion size estimation is an important source of error in obtaining dietary data. While food portion photos are known to aid the food portion estimation, the validation of photos are required. This study aimed at evaluating the bias and associated factors in quantifying the size of food portions eaten during lunch, and estimated by a 24-Hour Recall (24HR) interview, with and without the use of photos. Design and Participants: A validation study was carried out with 140 adults in the Brazilian cities of Curitiba and Aracaju. Data collection began in the first semester of 2017, lasting approximately one year. On the first day, participants consumed foods during a lunch with 10 possible items; each weighed after selection. The following day, a 24HR interview was applied. Participants were randomly allocated into two groups, according to the use of food portion photos during the interview. Means of the consumed and reported food amounts, and the difference between them were estimated. The effect of the use of photos was obtained by evaluating the odds of correctly estimating each food portion, (±10% and ±25%), using logistic regression. Additionally, the effect of using photos, the main predictor, was adjusted by other variables (e.g. age and educational level). Results: Rice, beans and meatballs had the greater differences between the assessed groups. The odds of success in estimating the portion size were greater amongst those that used the photo to estimate their consumption of rice and beans; while the contrary was observed for meatballs. Carrot, lettuce and juice had similar biases between the groups. Furthermore, educational level, cooking habit, and study centre location influenced the correct estimation of some foods. The evaluated photos can assist to estimate food consumption, especially two traditionally Brazilian foods, rice, and beans. The use of the meatball photo was unsatisfactory and further assessments are needed for the remaining photos. Thus, the evaluated photos album can be a useful tool in estimating Brazilian food consumption although further developments and assessments are also granted.

A comparison of food portion size estimation methods among 11\u201312\u2009year olds: 3D food models vs an online tool using food portion photos (Intake24)

BackgroundTechnology has advanced bringing new cost-effective methods to measure food intake. The aim of the study was to compare food and drink portion estimates from a traditional portion estimation method using 3D food models with portion estimates using an online dietary recall tool, Intake24.Methods11-12 year old children were recruited from secondary schools in Newcastle upon Tyne. Each pupil completed a two-day food diary followed by an interview during which pupils estimated food portion sizes using a range of 3D food models. They also completed Intake24 for the same 2 days. Bland Altman analyses were used to compare mean intake for each method.ResultsSeventy pupils completed both portion estimation methods. There was good agreement in food weight estimations between the two methods (geometric mean ratio 1.00), with limits of agreement ranging from minus 35% to plus 53%. Intake24 provided estimates of energy intake that were 1% lower on average than estimates of energy intake using the food models. Mean intakes of all macro and micronutrients using Intake24 were within 6% of the food model estimates.ConclusionsThe findings suggest that there was little difference in portion estimations from the two methods, allowing comparisons to be made between Intake24 data and food diary data collected from same age pupils using 3D food models in previous years.

An End-to-End Food Image Analysis System

Modern deep learning techniques have enabled advances in image-based dietary assessment such as food recognition and food portion size estimation. Valuable information on the types of foods and the amount consumed are crucial for prevention of many chronic diseases. However, existing methods for automated image-based food analysis are neither end-to-end nor are capable of processing multiple tasks (e.g., recognition and portion estimation) together, making it difficult to apply to real life applications. In this paper, we propose an image-based food analysis framework that integrates food localization, classification and portion size estimation. Our proposed framework is end-to-end, i.e., the input can be an arbitrary food image containing multiple food items and our system can localize each single food item with its corresponding predicted food type and portion size. We also improve the single food portion estimation by consolidating localization results with a food energy distribution map obtained by conditional GAN to generate a four-channel RGB-Distribution image. Our end-to-end framework is evaluated on a real life food image dataset collected from a nutrition feeding study.

Development and validation of a photographic atlas of food portions for accurate quantification of dietary intakes in China.

BackgroundAccurate estimation of food portion sizes remains an important challenge in dietary data collection. The present study aimed to develop a food atlas with adequate visual reference to improve the accuracy of dietary surveys in China.MethodsA food atlas for dietary surveys in China was developed using three visual reference systems, namely, regularly placed food portions, the two‐dimensional background coordinates and common objects known in daily life. The atlas was validated by estimating a meal before and after using the food atlas, and differences in weight estimation were compared using a paired t‐test. In total, 50 college students participated in the study.ResultsAfter determination of food varieties; design of the food display; purchase, processing, cooking and weighing of food; photographing food; post‐image processing and data processing, a total of 799 pictures of 303 types of food and two types of tableware were produced. The mean value of food weight estimated with the atlas was closer to the actual weight, and the variation range of these values was smaller and more stable than that estimated without the atlas. The differences estimated before and after using the atlas for all foods were significant (P < 0.05). Comparing the differences in weight before using the atlas, the error ranges of food samples were reduced.ConclusionsA food atlas has been developed for a retrospective dietary survey in China, which can be used to enable a better understanding of nutritional adequacy in the Chinese population.

A short education session increases the accuracy of estimated food records in young Korean women during a controlled-feeding study

BACKGROUND/OBJECTIVESDespite the widespread use of dietary assessment tools, the validity of food records has not been evaluated in Koreans. We assessed the accuracy of estimated food records and the effect of a short education session in young Korean women.SUBJECTS/METHODSThirty women (aged 18–23 yrs) each completed 3 food records during a controlled-feeding study. One educational session was provided on day 2 of the study. Food records were analyzed for the accuracy of food items and portion size estimation according to food group (grains; meat, fish, eggs, and beans; vegetables; fruit; dairy; and oils and sugars) and type of dish (rice, kimchi, soup, side dishes, spreads, beverages, and snacks). Reported food items were categorized as exact, close, or far matches, exclusions, or intrusions. Portion sizes were evaluated as accurate, similar, or inaccurate estimates, or missing. The means of days 2 and 3 were used to assess post-education results. Paired t-tests were performed to assess the effects of the education session.RESULTSThe mean percentages of exact matches, close matches, far matches, and exclusions on day 1 were 80.9%, 10.9%, 2.0%, and 6.2%, respectively, and mean intrusions observed were 0.1. The education session slightly increased the accuracy of recorded food items. The percentages of accurate, similar, and inaccurate estimates, and missing portion sizes were 11.7%, 19.8%, 12.2%, and 56.3%, respectively, at baseline. The percentage of missing portion size estimates decreased to 14.0% after the education session, resulting in an increase in the percentages of all other estimates. An increase was observed in the accuracy of reported portion sizes of vegetables, rice, and kimchi.CONCLUSIONSIn young Korean women, estimated food records are highly accurate for food items but not for portion size estimates without prior education. A short education session can improve the accuracy of portion size estimation.Trial RegistrationClinical Research Information Service Identifier: KCT0003307

A Systematic Review of Sensor-Based Methodologies for Food Portion Size Estimation

Food portion size estimation (FPSE) is critical in dietary assessment and energy intake estimation. Traditional methods such as visual estimation are now replaced by faster, more accurate sensor-based methods. This article presents a comprehensive review of the use of sensor methodologies for portion size estimation. The review was conducted using the PRISMA guidelines and full texts of 67 scientific articles were reviewed. The contributions of this article are three-fold: i) A taxonomy for sensor-based (SB) FPSE methods was identified, classifying the sensors (as wearable, portable and stationary) and the methodology (as direct and indirect). ii) A novel comprehensive review of the state-of-the-art SB-FPSE methods was conducted and 5 sensor modalities (Acoustic, Strain, Imaging, Weighing, and Motion sensors) were identified. iii) The accuracy of portion size estimation and the applicability to free-living conditions of these SB-FPSE methods were assessed. This article concludes with a discussion of challenges and future trends of SB-FPSE.

An evaluation of a virtual atlas of portion sizes (VAPS) mobile augmented reality for portion size estimation

Food portion size estimation is a critical yet challenging task in dietary assessment. Augmented reality technology enables the presentation of food dimensions and volume in a virtual three-dimensional object. It has the potential to improve perception and estimation of portion sizes. This study aims to develop and evaluate a novel mobile augmented reality application, namely Virtual Atlas of Portion Sizes (VAPS), as a portion size estimation aid. The development methodology of VAPS involves food photography, reconstruction of 3D models using photogrammetry method and presenting them in an AR environment. The 3D food models displayed in either semi-transparent or vivid mode for users to perform food portion estimation. Users can then resize and rotate the 3D models to fit the virtual model with the actual food. A total of thirty-six participants were involved in the evaluation and were divided into a health science and a non-health science background group. VAPS received good usability level with 76 SUS score. In terms of task completion time, unsurprisingly, the health science group performed faster. However, both groups have equivalent accuracy on the food portion estimation task using VAPS: 22.5% for non-health science group and 26.6% for health science group. The health science group liked and have better accuracy in vivid 3D food models (37.5%). Meanwhile, the non-health science group preferred semi-transparent 3D food models, but the accuracy is not significantly different between semi-transparent (25%) and vivid 3D food model (20%). Results demonstrate the potential of VAPS to aid in portion size estimation for dietary assessment, and participants’ feedback will be incorporated in the future for improvement of the app.

Review of the validity and feasibility of image-assisted methods for dietary assessment.

Accurately quantifying dietary intake is essential to understanding the effect of diet on health and evaluating the efficacy of dietary interventions. Self-report methods (e.g., food records) are frequently utilized despite evident inaccuracy of these methods at assessing energy and nutrient intake. Methods that assess food intake via images of foods have overcome many of the limitations of traditional self-report. In cafeteria settings, digital photography has proven to be unobtrusive and accurate and is the method of choice for assessing food provision, plate waste, and food intake. In free-living conditions, image capture of food selection and plate waste via the user's smartphone, is promising and can produce accurate energy intake estimates, though accuracy is not guaranteed. These methods foster (near) real-time transfer of data and eliminate the need for portion size estimation by the user since the food images are analyzed by trained raters. A limitation that remains, similar to self-report methods where participants must truthfully record all consumed foods, is intentional and/or unintentional underreporting of foods due to social desirability or forgetfulness. Methods that rely on passive image capture via wearable cameras are promising and aim to reduce user burden; however, only pilot data with limited validity are currently available and these methods remain obtrusive and cumbersome. To reduce analysis-related staff burden and to allow real-time feedback to the user, recent approaches have aimed to automate the analysis of food images. The technology to support automatic food recognition and portion size estimation is, however, still in its infancy and fully automated food intake assessment with acceptable precision not yet a reality. This review further evaluates the benefits and challenges of current image-assisted methods of food intake assessment and concludes that less burdensome methods are less accurate and that no current method is adequate in all settings.

OpenDRS: An Open-source 24-hour Recall for Mobile Devices (P13-004-19)

ObjectivesOur aim was to develop a 24-hour recall program using open-source language that can be readily adapted for use on multiple platforms and in different settings. MethodsWe developed the Open Dietary Recall System (OpenDRS), a multi-pass, 24-hour recall survey using the XLSForm programming language. XLSForm operates in Microsoft Excel and is used by several free or subscription-based electronic survey platforms that deploy forms via Android devices or online. The first pass collects a brief list of foods consumed, the second pass collects descriptive details and added ingredients for each food, and the third pass collects portion size estimates. The final pass reviews the recorded data with the option to edit or add foods. The program references external food and ingredient lists in .csv format. These food and ingredient lists are customizable for each survey setting, can be coded to match nutrient composition or recipe tables, and can be amended over the course of a study. Questions or response options can be added or edited to fit study-specific data collection methods. Photos can be incorporated for food identification or portion size estimation. Two case studies from adaptations in Malawi and Haiti will be presented. ResultsOpenDRS was used to collect 24-hour dietary intake data among 6- to 15-month-old children in Malawi, for a randomized, controlled complementary feeding trial. The median number of foods reported was 6, and the median survey duration was 9 minutes. For a national nutrition survey of 6- to 59-month-old children and their caregivers in Haiti, the program was expanded to record recipe data and to capture caregiver and child intakes in one electronic survey form. Recipes reported during the caregiver’s recall can be linked to the child’s recall if both consumed the same mixed dish. ConclusionsOpenDRS is an open-source 24-hour recall program which has been used in low- and middle-income countries. The XLSForm program for data collection and Stata code for data formatting will be made available to researchers conducting nutrition surveys via the Open Science Framework. Funding SourcesThe Bill & Melinda Gates Foundation and GAIN.

Validation of a Digitally Displayed Photographic Food Portion Size Estimation Aid Among Women in Urban and Rural Malawi (P13-012-19)

ObjectivesEasy-to-use tools to facilitate portion size estimation in low-income countries are needed. The objective of this study was to validate digitally displayed photographic portion size estimation aids (PSEAs) against a weighed meal record and compare findings with an atlas of printed photographic PSEAs and actual prepared food PSEAs in a low-income country. MethodsIn Blantyre and Chikwawa Districts, Malawi, we enrolled 300 women 18–45 years of age equally divided by urban/rural residence and years of education (≤4 years and >4 years). Participants served themselves water and five prepared foods, which were weighed separately before the meal and again after the meal to measure any leftovers. Participants returned the following day and completed a meal recall. They estimated the quantities of foods consumed three times using the different PSEAs in a randomized order. ResultsResponses for digital and printed PSEAs were highly correlated (>91% agreement for all foods, Cohen’s κ = 0.78–0.93). Overall, digital and actual-food PSEAs had a similar level of agreement with the weighed meal record at the group level. The proportion of participants who estimated within 20% of the weighed grams of food consumed ranged by type of food from 30–45% for digital PSEAs and 40–56% for actual food PSEAs. Digital PSEAs consistently underestimated grams and nutrients across foods, whereas actual-food PSEAs provided a mix of under- and overestimates that balanced each other to produce accurate mean energy and nutrient intake estimates. Results did not differ by urban and rural location or participant education level. ConclusionsDigital PSEAs require further testing in low-income settings to improve accuracy of estimations because they offer several logistical advantages over other PSEAs. Funding SourcesRTI International.

Feasibility of Reviewing Digital Food Images for Dietary Assessment among Nutrition Professionals.

Validity of image-assisted and image-based dietary assessment methods relies on the accuracy of portion size estimation based on food images. However, little is known on the ability of nutrition professionals in assessing dietary intake based on digital food images. This study aims to examine the ability of nutrition professionals in reviewing food images with regard to food item identification and portion size estimation. Thirty-eight nutritionists, dietitians, and nutrition researchers participated in this study. Through an online questionnaire, participants’ accuracy in identifying food items and estimating portion sizes of two sets of digital food images presenting a meal on a plate (Image PL) and in a bowl (Image BW) were tested. Participants reported higher accuracy in interpreting Image BW compared to Image PL, both in terms of accuracy in food identification (75.3 ± 17.6 vs. 68.9 ± 17.1%) and percentage difference in portion size estimation (44.3 ± 16.6 vs. 47.6 ± 21.2%). Weight of raw vegetables was significantly underestimated (−45.1 ± 22.8% vs. −21.2 ± 37.4%), while drink was significantly overestimated (40.1 ± 45.8% vs. 26.1 ± 32.2) in both images. Less than one-third of the participants estimated portion size within 10% of actual weight for Image PL (23.7%) and Image BW (32.3%). Accuracy of nutrition professionals in reviewing food images could be further improved with training on better perception of portion sizes from images.

Evaluation of a photographic food atlas as a tool for quantifying food portion size in the United Arab Emirates.

Although, United Arab Emirates (UAE) has one of the highest prevalence of overweight, obesity and type 2 diabetes in the world, however, validated dietary assessment aids to estimate food intake of individuals and populations in the UAE are currently lacking. We conducted two observational studies to evaluate the accuracy of a photographic food atlas which was developed as a tool for food portion size estimation in the UAE. The UAE Food Atlas presents eight portion sizes for each food. Study 1 involved portion size estimations of 13 food items consumed during the previous day. Study 2 involved portion size estimations of nine food items immediately after consumption. Differences between the food portion sizes estimated from the photographs and the weighed food portions (estimation error), as well as the percentage differences relative to the weighed food portion for each tested food item were calculated. Four of the evaluated food items were underestimated (by -8.9% to -18.4%), while nine were overestimated (by 9.5% to 90.9%) in Study 1. Moreover, there were significant differences between estimated and eaten food portions for eight food items (P<0.05). In Study 2, one food item was underestimated (-8.1%) while eight were overestimated (range 2.52% to 82.1%). Furthermore, there were significant differences between estimated and eaten food portions (P<0.05) for six food items. The limits of agreement between the estimated and consumed food portion size were wide indicating a large variability in food portion estimation errors. These reported findings highlight the need for further developments of the UAE Food Atlas to improve the accuracy of food portion size intake estimations in dietary assessments. Additionally, recalling food portions from the previous day did not seem to increase food portion estimation errors in this study.

Image-based food portion size estimation using a smartphone without a fiducial marker.

Current approaches to food volume estimation require the person to carry a fiducial marker (e.g. a checkerboard card), to be placed next to the food before taking a picture. This procedure is inconvenient and post-processing of the food picture is time-consuming and sometimes inaccurate. These problems keep people from using the smartphone for self-administered dietary assessment. The current bioengineering study presents a novel smartphone-based imaging approach to table-side estimation of food volume which overcomes current limitations. We present a new method for food volume estimation without a fiducial marker. Our mathematical model indicates that, using a special picture-taking strategy, the smartphone-based imaging system can be calibrated adequately if the physical length of the smartphone and the output of the motion sensor within the device are known. We also present and test a new virtual reality method for food volume estimation using the International Food Unit™ and a training process for error control. Our pilot study, with sixty-nine participants and fifteen foods, indicates that the fiducial-marker-free approach is valid and that the training improves estimation accuracy significantly (P0·05). Elimination of a fiducial marker and application of virtual reality, the International Food Unit™ and an automated training allowed quick food volume estimation and control of the estimation error. The estimated volume could be used to search a nutrient database and determine energy and nutrients in the diet.