TWM642885U - Computing device for evaluating diet - Google Patents

Abstract

A computing device for evaluating diet includes an output module, a storage module and a processing module for obtaining a food image related to a meal ingested by a user, and according to a food The recognition model identifies the at least one ingested food in the food image, and for each ingested food, calculates a reference ratio of the ingested food relative to a reference object, and obtains an ingested ratio, and obtains an ingested ratio For each ingested food, classify the ingested food into one of multiple food categories, for each food category, obtain the intake ratio corresponding to the food category, and calculate the intake ratio according to the intake ratio of each food category and the recommended intake ratio , obtain and output a diet suggestion output message through the output module.

Description

Computing device for evaluating diet

The present invention relates to a computing device for evaluating diet, in particular to a computing device for evaluating diet of a user.

According to statistics, 75% of the elderly in Taiwan suffer from at least one chronic disease, and more than 50% of the elderly suffer from two or more chronic diseases at the same time. These chronic diseases are often caused by bad habits or improper diet over a period of years. Bad eating habits can also lead to obesity. The scope and severity of the problems caused by obesity has gradually increased over the past few decades, and the treatment cost of related complications will reach 1.2 trillion US dollars. Early detection, intervention and treatment can effectively reduce the risk of complications and promote public health.

It can be seen that a balanced diet has an inseparable relationship with health. However, how to improve the dietary knowledge of the rural people in rural areas where professional medical care and nutritionists are insufficient, so as to promote the improvement of people's dietary concepts and establish healthy eating behaviors for Chinese people , and then to protect the health of the people of the country, it is necessary to find a solution.

Therefore, the purpose of the present invention is to provide a computing device for evaluating diet that can automatically evaluate the diet status of the user anytime and anywhere.

Therefore, the computing device for evaluating diet of the present invention includes an output module, a storage module and a processing module electrically connected to the output module and the storage module.

The storage module stores a food identification model for identifying at least one food included in an image.

The processing module is used to obtain a food image related to a meal ingested by the user, the food image includes at least one ingested food and a reference object, and the at least one food image in the food image is identified according to the food recognition model A food intake, for each food intake, calculate the reference ratio of the food intake relative to the reference object, and obtain the actual intake of the food intake of the user relative to the reference object according to the user's input operation An intake ratio of the ingested food, and obtain an intake ratio according to the reference ratio and the intake ratio corresponding to the ingested food, and classify the ingested food into one of multiple food categories, for each For the food category, sum up the intake proportions corresponding to the intake foods corresponding to the food category and divide it by the sum of all intake proportions to obtain the intake proportion corresponding to the food category, and calculate the intake proportion of each food category and each A suggested intake ratio of a food category is obtained and a dietary suggestion output message related to the user's diet is output through the output module.

The effect of the present invention is: the food image is recognized by the computing device The at least one ingested food, and calculate the intake ratio corresponding to each ingested food, and classify each ingested food into one of multiple food categories, and obtain the intake ratio corresponding to each food category, and according to The intake ratio of each food category and the recommended intake ratio of each food category, obtain and output the output information of the dietary suggestion, so as to achieve the effect of automatically evaluating the user's dietary status anytime and anywhere.

1: computing device

11: Output module

12: Communication module

13: Storage module

14: Processing module

100: Communication network

101:Evaluation side

21: Steps

211~213: Steps

31~39: Steps

Other features and functions of the present invention will be clearly presented in the embodiments with reference to the drawings, wherein: Fig. 1 is a block diagram illustrating an embodiment of the present invention's computing device for evaluating diet via a communication network and a Evaluation terminal is connected; Fig. 2 is a flow chart, illustrates and utilizes the embodiment of the computing device of the present invention to be used for evaluating diet and realizes a food identification model obtaining program of diet evaluation method; Fig. 3 is a flow chart, illustrates how this computing device Establishing a food recognition model; and FIG. 4 is a flow chart illustrating a diet evaluation procedure for realizing the diet evaluation method using the embodiment of the computing device for evaluating diets of the present invention.

Referring to Fig. 1, the embodiment of the computing device 1 of the present invention is used for evaluating diet is connected with an evaluation terminal 101 through a communication network 100, and is suitable for evaluating a user's diet status, and comprises an output module 11, a connection communication to the communication network 100 The module 12 , a storage module 13 and a processing module 14 electrically connected to the output module 11 , the communication module 12 and the storage module 13 . Implementations of the computing device 1 and the evaluation terminal 101 are, for example, a server, a personal computer, a notebook computer, a tablet computer, or a smart phone.

The storage module 13 stores multiple sets of training data, each set of training data includes a training food image including at least one training food, and each training food image marks the location and food type of the corresponding training food. In this embodiment, each training food image uses a bounding box to mark the position of the corresponding training food.

The operation details of the computing device 1 will be described below through a diet evaluation method, which sequentially includes a food recognition model obtaining program and a diet evaluation program.

Referring to FIG. 1 and FIG. 2 , the food recognition model obtaining program illustrates how to obtain a food recognition model for recognizing at least one food included in an image, and includes the following steps.

In step 21 , the processing module 14 uses an object detection deep learning method such as YOLOv4 to build the food recognition model according to the training data stored in the storage module 13 .

It is worth noting that step 21 includes the following sub-steps (see FIG. 3 ).

In step 211, the processing module 14 performs data enhancement processing on the training data. In this embodiment, the processing module 14 uses a Mosaic data Enhancement method to perform data enhancement processing.

In sub-step 212, the processing module 14 trains the network structure of a YOLO model according to the training data processed in step 211, wherein the network structure of the YOLO model includes a backbone network, a connection structure and an output layer. In this embodiment, the processing module 14 uses CSPdarknet53 as the backbone network, and uses the spatial pyramid pooling and path aggregation network as the connection structure.

In sub-step 213 , the processing module 14 performs repeated training until the loss function of the YOLO model converges to obtain the food recognition model.

Referring to Fig. 1 and Fig. 4, the diet evaluation program illustrates how to automatically evaluate a user's diet status, and includes the following steps.

In step 31 , the processing module 14 obtains a food image related to a meal ingested by the user, and the food image includes at least one ingested food and a reference object. It is worth mentioning that when shooting the food image, the at least one ingested food should avoid overlapping and covering each other as much as possible, and try to make the paving height of each ingested food consistent, so as to facilitate subsequent identification processing.

In step 32, the processing module 14 identifies the at least one ingested food in the food image according to the food identification model.

In step 33, for each ingested food, the processing module 14 calculates a reference ratio of the ingested food relative to the reference object. It is worth mentioning that, in this embodiment, the reference object can be a fist of the user, and the processing module 14 The reference object is first identified and the bounding box of the reference object in the food image is framed. Then, the processing module 14 calculates the ratio of the ingested food in the bounding box of the food image to the bounding box of the reference object to obtain The reference ratio. By calculating the ratio of the ingested food to the user's fist, the amount of each ingested food can be estimated. For example, the at least one ingested food is, for example, rice, ribs, and cabbage. If the ratio of rice to the user's fist is 1, it is estimated that the amount of rice is about the amount of one fist of the user, and the ratio of ribs to the user's fist is estimated to be 1. If the ratio of the user's fist is 1/2, it is estimated that the weight of ribs is about half the weight of the user's fist. If the ratio of cabbage to the user's fist is 1/3, it is estimated that the weight of cabbage is about 1/3 of the user's fist.

In step 34, for each ingested food, the processing module 14 obtains an intake ratio of the user's actual intake of the ingested food relative to the ingested food according to the user's input operation.

In step 35, for each ingested food, the processing module 14 obtains an ingestion ratio according to the reference ratio and the ingestion ratio corresponding to the ingested food, and sends the A diet data including the intake ratio of each food intake is sent to the evaluation terminal 101 . Thereby, the ingestion amount of the ingested food that the user actually ingests can be estimated. Continuing the above example to illustrate, if the proportion of rice intake is 2/3, it is estimated that the actual amount of rice consumed by the user is about 2/3 of the user's fist. is 1, it is estimated that the user actually The amount of pork ribs actually ingested by the user is about half the amount of the user's fist. If the intake of cabbage is 1/2, it is estimated that the actual amount of cabbage ingested by the user is about 1/6 of the user The weight of a fist.

In step 36, for each food intake, the processing module 14 classifies the food intake into one of a plurality of food categories. The food categories include whole grains, protein, vegetables and fruits.

In step 37, for each food category, the processing module 14 sums the intake ratios corresponding to the ingested foods corresponding to the food category and divides it by the sum of all intake ratios to obtain the intake corresponding to the food category Proportion. In this way, the intake ratio of the user's actual intake of the food category can be automatically estimated, instead of simply estimating the intake of each food intake, and the eating status of the user can be further checked. Continuing the above example to illustrate, the amount of rice that the user actually ingests is about 2/3 of the user's fist. 1/2, the amount of pork ribs actually ingested by the user is about half the amount of the user's fist, ribs belong to the protein, so the intake ratio of the protein that the user actually ingests is 3/8, the user The amount of cabbage actually ingested by the user is about 1/6 of the fist of the user. Cabbage belongs to this vegetable, so the intake ratio of the vegetable actually ingested by the user is 1/8. In addition, this meal should use The user did not ingest fruit, so the ingestion ratio of the fruit actually ingested by the user was 0.

In step 38, the processing module 14 obtains and outputs a dietary suggestion output message related to the user's diet through the output module 11 according to the intake ratio of each food category and the recommended intake ratio of each food category . In this embodiment, if the dietary guideline recommended by the Healthy Diet Plate published by Harvard Health Magazine is used as the recommended intake ratio, the recommended intake ratio of the whole grains is 1/4, and the recommended intake ratio of the protein is 1/4. 1/4 and the total recommended intake ratio of the vegetables and fruits is 1/2. This is compared with the user's intake of whole grains as 1/2, protein as 3/8, and vegetables and fruits as 1/8, the processing module 14 will The diet suggestion output message indicating that the user has too much intake of whole grains and insufficient intake of vegetables and fruits is automatically output.

In step 39 , after the processing module 14 receives an evaluation result from the evaluation terminal 101 in response to the dietary data through the communication module 12 via the communication network 100 , the evaluation result is presented. Among them, in addition to automatically outputting the diet suggestion output message, the user's diet data can also be sent to the evaluation terminal 101 held by professional doctors and nutritionists, and professional doctors and nutritionists can provide further evaluation As a result, the user can further strengthen his diet concept.

To sum up, the present computing device for evaluating diet uses the computing device 1 to identify the at least one ingested food in the food image, and calculates the intake corresponding to each ingested food actually consumed by the user. proportion, rather than just calculating the reference proportion corresponding to each ingested food in the meal ingested by the user, because the user It may not be possible to eat every ingested food in the meal without being picky eaters at all, so there is still a difference between the reference ratio corresponding to each ingested food and the actual ingested ratio of the user. Therefore, by calculating the corresponding ratio of each ingested food The proportion of intake can make the follow-up evaluation more accurate, and then, the calculation device 1 classifies each food intake into one of multiple food categories, and obtains the intake proportion corresponding to each food category, and according to each food The intake ratio of each food category and the recommended intake ratio of each food category are obtained and output as the dietary suggestion output information, so as to achieve the effect of automatically evaluating the user's dietary status anytime and anywhere, so the purpose of the new model can indeed be achieved.

But the above-mentioned ones are only embodiments of the present invention, and should not limit the scope of implementation of the present invention with this. All simple equivalent changes and modifications made according to the patent scope of the present application and the content of the patent specification are still within the scope of the present invention. Within the scope covered by this patent.

1: computing device

11: Output module

12: Communication module

13: Storage module

14: Processing module

100: Communication network

101:Evaluation end

Claims (5)

A computing device for evaluating diet, suitable for evaluating a user's diet status, comprising: an output module; a storage module storing a food identification model for identifying at least one food included in an image; and A processing module, electrically connected to the output module and the storage module, and used to obtain a food image related to a meal ingested by the user, the food image includes at least one ingested food and a reference object, and according to The food recognition model identifies the at least one ingested food in the food image, and for each ingested food, calculates a reference ratio of the ingested food relative to the reference object, and obtains the ingested food according to the user's input operation. The actual intake of the ingested food by the user is compared to an intake ratio of the ingested food, and an intake ratio is obtained based on the reference ratio and the intake ratio corresponding to the ingested food, and the ingested food is If it is classified as one of multiple food categories, for each food category, sum up the intake proportions corresponding to the intake foods corresponding to the food category and divide by the sum of all intake proportions to obtain the intake corresponding to the food category ratio, and according to the intake ratio of each food category and the recommended intake ratio of each food category, obtain and output a diet suggestion output message related to the user's diet through the output module. The computing device for evaluating diet as described in Claim 1, wherein the storage module also stores multiple sets of training data, each set of training data includes a training food image containing at least one training food, and each training food image The location and food type of the corresponding training food are marked, and the processing module also uses an object detection deep learning method to establish the food recognition model based on the training data. The computing device for evaluating diet as described in claim 2, wherein the processing module performs data enhancement processing on the training data, and trains a network of YOLO models according to the enhanced training data structure, wherein the network structure of the YOLO model includes a backbone network, a connection structure and an output layer, and repeated training is performed until the loss function of the YOLO model converges to obtain the food recognition model. The computing device for evaluating diet as described in claim 3, wherein the processing module uses a Mosaic data enhancement method to perform data enhancement processing, and uses CSPdarknet53 as the backbone network, and uses spatial pyramid pooling and routing Converged networks serve as connectivity structures. The computing device for evaluating diet as described in claim 1 is also connected to an evaluation terminal via a communication network, and the computing device also includes a communication module electrically connected to the processing module and connected to the communication network, the After the processing module receives an evaluation result from the evaluation terminal in response to the diet data via the communication module, it presents the evaluation result.

Images