no health problem
Conditions
Interventions
Group 1: Thirty participants document their food intake over four days using standardized weighing protocols and photographic records from two perspectives
Sponsors
Technische Universität München
Eligibility
Sex/Gender
All
Age
18 Years to No maximum
Inclusion criteria
Inclusion criteria: Written informed consent
Exclusion criteria
Exclusion criteria: Subjects under 18 years of age. Lack of informed consent Withdrawal of consent during the study
Design outcomes
Primary
| Measure | Time frame |
|---|---|
| The agreement between the nutritional evaluation based on manual input (text) and automated image recognition | — |
Secondary
| Measure | Time frame |
|---|---|
| a) Accuracy of image recognition depends on meal complexity. Image Type 1 (simple meal): Single ingredients, such as an apple or a portion of rice. Image Type 2 (complex meal): Composite or processed foods, e.g., a mixed salad. For Image Type 2, it is examined whether additional text input is necessary to improve nutrient estimation, as certain properties (e.g., fat content of milk) are not visually detectable. b) Image Analysis Comparison: Meal photos are analyzed using XcellentDiet and other nutrition apps to compare nutrient outputs. Since software like OptiDiet does not support image analysis, it serves as a reference for text-based inputs. c) Consistence of Image Recognition: Each image is analyzed ten times to assess the repeatability of the AI detection. Photos are also taken from various angles especially for simple foods (e.g., apple), to evaluate the impact on 3D reconstruction and recognition accuracy. Camera parameters are also considered. d) Validation of Voice Input: A food item is entered into the app by several people using voice input. Speech recognition is checked for repeatability and accuracy. | — |
Countries
Germany
Contacts
Public ContactTianxing Du
Technische Universität München
Outcome results
None listed