ChewNet: Dataset for Invivo and Invitro Beef and Plant-based Burger Patty Boluses
Description
This dataset examines the chewing dynamics of beef and plant-based burger patties using both human (InVivo) and robotic (InVitro) methods, aiming to (1) identify the optimal robotic chewing cycles that mimic human swallowing and (2) extract bolus texture properties via deep learning-based image analysis. For in vivo trials, three healthy male participants provided near-swallow bolus samples, imaged with a 12MP camera and flatbed scanner, followed by Texture Profile Analysis (TPA). In vitro tests used a 3-DOF chewing robot (ChewBot) with adjustable molar trajectories, artificial saliva (10% food weight), and up to 40 chewing cycles with images, force profiles (100ms intervals), and TPA metrics (after every 5 cycles). ChewNet Dataset is organised into InVivo and InVitro folders, which are crucial for food science research (optimising texture in meat analogues), development of artificial mastication systems for food image processing tasks, and analysing in vitro food bolus properties at the swallowing stage.
Files
Steps to reproduce
Food Sample Preparation Two commercially available products tested in this research are the Impossible Plant-Based Burger Patties [1] and Silver Fern Farms Pure 97% Beef Sliders With Brisket [2]. These burger patties were cooked in the oven at 200°C until the temperature in the centre reached 74°C. The cooked patties were cooled down to room temperature. Prior to the test, the patties were cut into small pieces or also known as samples, of roughly 1 cm × 1 cm × 1.5 cm and weighing around 2g each for consistency. InVivo Chewing Experiments Three male participants (aged 29) were first instructed to naturally chew and swallow three samples of each food type (beef and plant-based) using one side of the mouth. The chewing process was recorded to determine the number of chewing cycles before swallowing. Subsequently, another three samples of each food type were provided, which participants chewed but spat out just before swallowing. Required ethics approval was obtained by the University of Auckland Human Participants Ethics Committee (Ethics Reference: UAHPEC26909). InVivo Post-Chew Data The expectorated food boluses were collected, weighed, and imaged using both a 12MP camera and a flatbed scanner. TPA was then performed to assess key textural parameters of the bolus at the point of near-swallowing. The acquired images and TPA data are available in the InVivo Data folder. InVitro Chewing Experiments (ChewBot) InVitro chewed food boluses were collected using the biomimicking chewing robot with 3-DOF (degree-of-freedom) equipped with a soft robotic oral cavity [3]. • Tested trajectories: T1 (minimal shear) and T13 (maximal shear) • Chewing cycle duration: 3 seconds • Chewing cycles: 1–40 in increments of 5 (5, 10, 15, ..., 40) • Saliva volume: 0.2 mL (10% of food weight) • One sample is fed to the robot per chewing test • Replicates: 5 repetitions per experiment setting InVitro Post-Chew Data • An image of the chewed food bolus captured as soon as the robotic jaw opens up. • Robot's chewing force profiles (Newtons) recorded every 100ms during each chewing cycle. • TPA metrics for chewed food boluses after each 5, 10, 15, 20, 25, 30, 35, and 40 chewing cycles.
Institutions
Categories
Funding
Riddet Institute