Operation support of kitchen appliance

Abstract

The invention discloses a kitchen appliance for processing food materials. The appliance comprises a container (120) for containing the food materials; a drive unit (125) with a mechanical tool (130) for handling the food material; input means (115, 160) for determining a desired state of the food material; a camera (135) for capturing an image of an object contained in the container (120); processing means (140) for determining the state of the food material on the basis of the image; and output means (115, 160) for outputting indication information of the determined state of the food material.

Description

TECHNICAL FIELD

[0001] The invention relates to a kitchen appliance. More specifically, the invention relates to the operation of said kitchen appliance during the preparation of a dish. BACKGROUND

[0002] A kitchen appliance is intended to support a user in the preparation of a dish. For example, the appliance can comprise a certain blending device for stirring, mixing or kneading ingredients in a container. The user can place the ingredients required for the dish into the container and process them by means of the kitchen appliance. Depending on the intended use of the ingredients in the dish or the desired effect to be achieved, different processing approaches can be required. Furthermore, the quality, freshness or temperature of the ingredients can vary, so that the processing approach can need to be adjusted.

[0003] Although there are related approaches that provide support in the preparation of a specific dish by a user, there is still a lack of advanced solutions that provide dynamic or individualized support for the operation of a kitchen appliance. It is therefore an object of the invention to provide a better way of controlling a kitchen appliance that can be operated by a user to process ingredients required for a dish. SUMMARY

[0004] The invention solves the above-mentioned problems by the technical solutions of the independent claims. The dependent claims define preferred embodiments.

[0005] According to a first aspect of the invention, a kitchen appliance for processing ingredients is proposed. The appliance comprises a container for accommodating ingredients, a drive unit with a mechanical tool for handling the ingredients, input means for determining a desired state of the ingredients, a camera for taking an image of the content of the container, processing means for determining the state of the ingredients based on the image, and output means for outputting an indication of the determined state of the ingredients. The state can be in particular a processing state, a handling state or a process state. All these types of states are to be understood as having the same meaning, unless explicitly stated otherwise.

[0006] The appliance can comprise a food processor, a blender or a similar device. By employing computer vision techniques, the result of the mechanical processing of the ingredients can be recognized, so that the processing can be controlled more precisely. The processing state of the ingredients can be described for example by a preset scale, which can range from "not processed" to "over-processed", and the desired processing state can correspond to any point on this scale.

[0007] It has been found that the mechanical processing of the ingredients can have a significant influence on the actual processing result. Conventional processing parameters, such as a speed setting and a processing time target, can not reliably achieve the desired effect, since the quality of the ingredients can be unknown and unforeseen circumstances can also influence the processing result.

[0008] The present application facilitates a more precise and reliable processing of foodstuff to achieve a desired processing result. The foodstuff can be subjected to frequent or continuous optical evaluation to achieve an optimal processing result. Thus, the kitchen appliance can simulate a quality assurance process of a chef. The determination of the processing state of the foodstuff can be independent of the previous processing duration, so that the state detection can be resumed smoothly even if the processing is interrupted.

[0009] By employing the present application, even a delicate process such as the preparation of mayonnaise can be completed without risk of failure, while at the same time over-processing of the foodstuff is avoided. If the foodstuff does not reach the desired state, a warning can be issued. The desired processing state can be selected from a wide range, so that the effect to which the processing result should be defined can be precisely defined. In some embodiments, the desired processing state can comprise a plurality of parameters. For example, when preparing a smoothie, the added fruit should be completely blended, and the smoothie should reach a predefined viscosity.

[0010] The "foodstuff" referred to herein can also include intermediate products in the preparation of a dish. For example, the foodstuff can include cream, yogurt or fruit when preparing a dessert, and the processing of one or more foodstuffs can result in the dessert. For example, a cake dough can be processed in the kitchen appliance, and the dough has to be baked after processing to produce a cake, so that the kitchen appliance can be used for the preparation of intermediate products of a dish.

[0011] The indication information can include a notification that the foodstuff has reached the desired processing state. Upon receipt of the notification, the user can stop the processing, change the processing settings or add further foodstuffs for processing. The notification can in particular be output to the user in the form of sound, vision or haptics.

[0012] In some embodiments, the appliance can be controlled using the indication information. More specifically, the processing device can be adapted to control the drive unit such that the foodstuff is processed to reach the desired (processing) state. It is further preferred that the drive unit is controlled such that the foodstuff reaches the desired (processing) state. This can involve adjusting processing parameters such as the processing speed, the direction of movement or the type of movement imparted to the mechanical tool. A series of adjustments to the processing parameters can be made, for example, a rapid stirring of the foodstuff, followed by a slow stirring for a certain period of time, followed by a powerful stirring. In some embodiments, a plurality of parameters can be adjusted simultaneously to control the processing of the foodstuff. For example, the mechanical tool can comprise a rotatable kneading hook, and the appliance can be adapted to rotate the container independently of the kneading hook.

[0013] The desired processing state of the foodstuff can be determined on the basis of a recipe for preparing a particular dish. Thus, the indication information of the desired processing state can be understood in the context of the overall goal intended, which facilitates a clear definition of the specific requirements of the desired state, so that the state of the foodstuff after processing is more in line with the requirements in the context of the preparation of the dish. For example, the same stirring of the foodstuff can be required for different reasons when preparing mashed potatoes and when preparing a smoothie.

[0014] The kitchen appliance can further comprise a further sensor for detecting a content of the container; the processing device can determine the processing state of the foodstuff in combination with a quantity detected by the further sensor. The further sensor can be configured to detect a weight, a temperature, a humidity, a texture or an odor of the foodstuff in the container. It is particularly preferred to use a plurality of sensors to detect a plurality of aspects of the foodstuff during processing, so that the processing state can be determined more accurately, more quickly or more reliably by taking into account the detection data of the plurality of sensors.

[0015] The processing device can determine the processing state of the foodstuff using a machine learning technique. The machine learning technique can in particular comprise an approach known as artificial intelligence (AI), for example using a multi-layer artificial neural network (ANN) which can be trained by deep learning. It is preferred that the artificial neural network is trained by inputting known combinations of sensor data and corresponding processing states, which can also include an identification of the foodstuff being processed. Furthermore, it can also include information about the environmental conditions at the time of data acquisition, for example the time of day, the type of kitchen appliance used and / or the settings of the appliance at that time.

[0016] The machine learning technique is preferably based on pre-processed data. For example, images of the foodstuff in the container can be corrected for distortion, de-noised, compensated for lighting conditions or a region of interest (ROI) can be determined. The technique can also be used to predict the sensor data when the foodstuff reaches a desired processing state, and then compare the actual sensor data with the predicted data to determine how far the foodstuff is from reaching the desired processing state.

[0017] The processing device can train the artificial neural network using the machine learning technique based on the observed data of the foodstuff during processing. The training of the artificial neural network can be performed by the processing device locally on the kitchen appliance; in some embodiments, the training can also be performed on another device, in particular a device external to the kitchen appliance. The training can require a large data set and can involve computationally intensive operations, so the device used for the training can be equipped with a large data storage and / or a high-performance processing unit.

[0018] The kitchen appliance can be adapted to collect observation data during the food material processing. Take whipping cream as an example: when whipping cream continuously, the cream goes through different stages or processing states, which can be observed by the camera and possibly other sensors. The first stage can be liquid cream with foam, the second stage is very soft peak state, the third stage is soft peak state, the fourth stage is hard peak state, the fifth stage is over-whipped cream (basically cannot be used for food making), and the sixth stage is butter. If a certain amount of liquid cream is processed to, for example, the fourth stage, the change process of the cream from the first stage to the fourth stage can be observed. The observed data can be used to further train the artificial neural network by machine learning technology. The user can express the satisfaction degree of the achieved processing state, and the satisfaction feedback can be used to add labels to the training data.

[0019] The kitchen appliance can also include a communication device, especially for exchanging data with a remote computing backend. The remote computing backend can provide a trained machine learning model for performing the above-mentioned determination of the food material processing state. The kitchen appliance can provide the observed data to the remote computing backend, and the backend can train the model using the data.

[0020] According to another aspect of the application, a system is provided. The system includes the above-mentioned kitchen appliance, and a computing backend running a food processing base model. The computing backend is adapted to receive, from the kitchen appliance, at least one detected parameter of the content in the container, indication information that the food material is to be processed, and user indication information for operating the household appliance; The base model provides the parameters required by the kitchen appliance to perform the processing process.

[0021] Unlike existing support models, the parameters or settings of the kitchen appliance can be determined in combination with user factors. Cooking is a complex activity, and different users may have different perceptions of dishes (i.e. cooking results). By considering the user operating the kitchen appliance, the processing process can be controlled to ensure that the user is satisfied with the processing result.

[0022] The food making preferences of the user can be associated and the above-mentioned parameters can be determined based on the preferences. The preferences can include taste, food texture, crispness, cooking style, or food properties obtained by processing. In some embodiments, the user can describe his or her preferences in the form of user settings, etc.; in other embodiments, the user's preferences can be determined by observing the user. User preferences can include, for example, the selection of a dish making recipe, or other personal information such as race, preferred diet type, or food materials to be avoided.

[0023] Other available information related to the user can also be considered, which can be used to associate the user with a group of users that are similar in some aspects, and in turn apply the collective preferences of the group to the user. If a presumed preference is proven to be inaccurate, it can be corrected by removing the user from the original group and possibly associating him to another group.

[0024] The user's food preparation behavior can be associated and the above parameters determined based on the behavior. The behavior data can include the user's adjustments to recipes or cooking habits, such as the user's proficiency in cooking dishes or frequently occurring ingredient combinations.

[0025] The above parameters are preferably determined based on the dish to be prepared. In this scenario, the ingredients processed in the kitchen appliance are components of the dish, and thus changes in the way the ingredients are processed can affect the characteristics of the final dish. BRIEF DESCRIPTION OF DRAWINGS

[0026] Non-limiting embodiments of the present application will be described in detail in the following, with reference to the attached drawings: Figure 1 A system is shown; and Figure 2 An exemplary flowchart 200 for controlling a kitchen appliance 105 is shown. DETAILED DESCRIPTION

[0027] Figure 1 An exemplary system 100 is shown, which comprises a kitchen appliance 105, a computing backend 110 and optionally a mobile device 115. The kitchen appliance 105 shown in the figure is a food processor, but other types of appliances 105 can also be used.

[0028] The appliance 105 comprises a container 120 for holding the ingredients of a dish, a drive unit 125 with a mechanical tool 130 to handle the contents of the container, a camera 135 and a processing device 140. Optionally, one or more additional sensors can also be provided, such as a weight sensor 145, a temperature sensor 150 or a VOC (volatile organic compound) sensor 155. The processing device 140 can be connected to a communication unit 152. A user interface 160 can be operated by the user, and optionally the mobile device 115 can also be used as a user interface.

[0029] The container 120 typically comprises a bowl, which can be made of plastic, metal or glass. Foodstuff can be placed in the container 120 and processed by a mechanical tool 130. The mechanical tool 130 can for example be a kneading hook, a rotating blade or a ball-shaped stirring net. The drive unit 125 typically comprises an electric motor, possibly also a fixed or controllable reduction gear. The kitchen appliance 105 can support a plurality of different mechanical tools 130, which can be mounted to and removed from the drive unit 125 by the user. The drive unit 125 and / or the mechanical tool 130 can be configured for a specific task, for example by selecting an appropriate reduction ratio or assembling a plurality of mechanical tools 130 together.

[0030] The camera 135 is preferably arranged above the container 120, in particular when the container does not need to be covered during processing, or the lid of the container 120 is transparent, or an opening is provided in the lid for the camera 135 to take pictures. If the container 120 is made of a transparent material, the camera 135 can also be mounted on the side of the container 120.

[0031] The camera 135 is preferably capable of taking color images of the contents of the container 120, in particular preferably during processing without interrupting the processing. The camera 135 can work in the visible spectral range. Optionally, an illumination device can be integrated, ensuring that the lighting conditions are known, controllable or adjustable when taking images of the contents of the container.

[0032] The weight sensor 145 is adapted to determine the weight of the container 120 and its contents, and by observing the change in weight, the amount of foodstuff added to or removed from the container 120 can be determined. The temperature sensor 150 can be adapted to determine the temperature of the container; in some embodiments, a plurality of temperature sensors 150 can be provided at different locations of the container 120, such as near the bottom and on the side.

[0033] The VOC sensor 155 is used to detect volatile organic compounds, which are typically associated with odors, and can detect different odors emitted by the foodstuff in the container.

[0034] Figure 1 Further sensors, not shown, can be present or can differ. In some embodiments, a humidity sensor can be provided to determine the humidity of the foodstuff; also a current sensor can be provided to determine the current flowing through the drive unit 125.

[0035] The kitchen appliance 105 can be controlled by a user interface 160, which can be used to receive user input and / or output information to the user. The input can for example be in tactile or acoustic form, and the output can include acoustic, visual or tactile information. The mobile device 115 can be connected to the kitchen appliance 105 by communication means and support corresponding user interface functionality; preferably, a wireless connection is used, such as Bluetooth or Wi-Fi.

[0036] The user interface 160 can also be adapted to provide user feedback to the kitchen appliance 105, for example, the user's satisfaction with the result of a certain processing. The user can also select or view cooking recipes through the interface 160, the selected recipe can contain food material processing information, and the user can perform the corresponding processing procedure through the kitchen appliance 105.

[0037] In some embodiments, the processing device 140 is adapted to implement or run an artificial intelligence AI model that determines the processing state of the food material in the container 120 based on the detection data of the camera 135 and optionally one or more additional sensors 145-155. During the processing, the user can be outputted with an indication of the current processing state.

[0038] The user can set the desired processing state of the food material through the user interface 160. The processing device 140 can indicate the gap between the current processing state and the desired state, including the estimated remaining processing time. The processing device 140 can also be adapted to control the household appliance 105, in particular the drive unit 125, so that the food material in the container 120 is processed to reach the desired state; when the food material reaches the state, the processing can be stopped or maintained at a low intensity.

[0039] The computing backend 110 can be adapted to train an artificial intelligence AI model that determines the processing state of the food material based on camera images and possibly more sensor data. The trained model can be downloaded into the kitchen appliance 105 for use or implementation by the processing device 140.

[0040] According to another aspect of the present application, the processing device 140 and / or the computing backend 110 can provide control parameters for controlling the kitchen appliance 105 to process a predetermined food material. The parameters can be determined according to the indication of the food material to be executed processing procedure and the indication of the user operating the household appliance 105, and can also consider the indication of the processing desired result. The household appliance 105 can then be controlled based on the parameters.

[0041] Figure 2 An exemplary flowchart 200 for controlling the kitchen appliance 105 is shown. The flowchart is based on the kitchen appliance 105 discussed above in connection with Figure 1 The software running on the kitchen appliance 105 uses known data processing techniques to process sensor data. The software may, for example, include: - Real-time computer vision, image processing and / or machine learning library toolboxes programming functions (such as Python and / or C language) for implementing interoperability in various integrated development environments; - Computer vision tools: OpenCV; - Image processing tools: keras, tensorflow, pytorch, numpy, image labeler and object detection feature generator; - Inference architecture: Artificial Neural Network (ANN) model; - Statistical methods: Time series regression and discrete classification (MobileNet, Tensorflow Lite).

[0042] In step 205, sensor data acquisition is triggered. When a trigger event is detected, an image is captured by the RGB image sensor in the camera 135. Possible trigger events include: - Machine power button activation; - Mechanical tool positioning detection; - Weight measurement by a load cell; and / or - Graphical user interface (GUI) / mobile application virtual start button trigger.

[0043] The latest image signal can be tagged in ascending order or time stamp.

[0044] In step 210, sensor data is acquired. The following operations can be performed, for example: - Transfer of images from device (sensor device) to host (microprocessor) via wired connection (Ethernet RJ45, I2C, SPI) or wireless transmission (Wi-Fi and / or Secure Shell (ssh) connection); - Perform image reading, writing and display using OpenCV functions (imread, imwrite, imshow) from the AL library toolbox; - Perform sensor and temperature calibration using software scaling factors or adjustment methods if necessary; - Use parallel computing toolbox to take advantage of multi-core processors to shorten data processing time; - Adjust sensor position to ensure that image capture is not obstructed: delay image capture after detecting sensor positioning activation; - Improve image brightness by LED light emission in low light conditions; - Use RGB technology without infrared filter to reduce reflection effects from the container 120, mechanical tool and food ingredients.

[0045] In step 215, statistical analysis is performed on the acquired data. Step 215 can include, for example, the following operations: - Apply pre-processing techniques to remove noise from images: blurring, sliding window processing, brightness adjustment, segmentation, etc. - Apply statistical pre-processing techniques to remove noise and extract features from time series inputs: mean calculation, kurtosis calculation, skewness calculation, sliding window domain transformation (e.g. wavelet transform and Fourier series).

[0046] In step 220, an artificial intelligence, AI, model for food preparation reasoning is trained.

[0047] - A complete food preparation process is performed to collect a dataset; - The collected dataset, including images and time series signals (weight, current, temperature), is used for model training; - Feature extraction and selection are performed to select the most likely input combination, avoiding model overfitting problems; - The prediction model generates a prediction output in the form of a region of interest (ROI) and / or a classification name (class number or synthetic name label); - The accuracy of the prediction model is tested through new food preparation trial runs; - Example: A whipped cream image dataset for making coffee topping is collected for model training. The cream starts from a liquid state, goes through soft peak, hard peak, under-whipped, and over-whipped states, and finally becomes butter. Each stage produces unique shape, surface texture, and / or waveform features for pattern recognition. An artificial intelligence, AI, model can be trained to recognize the state transition stages when receiving new images and / or time series signals as input combinations.

[0048] In step 225, an output is provided: - The user interface 160 can support real-time monitoring of the food preparation process; - If temporary adjustments are needed, a pause and resume function can be enabled; after resuming operation, the food preparation progress is recalculated; then the user interface 160 can display the estimated remaining time to complete the scheduled food preparation process.

[0049] In step 230, process parameters are controlled: - The drive unit 125 is controlled to start the processing process of the kitchen appliance 105 on the food material; - When the determined processing state is consistent with the expected processing state, the drive unit 125 stops processing; - After resetting, the mechanical tools of the kitchen appliance return to the initial state, and the food preparation process and progress bar return to the 0 stage.

[0050] In step 235, user feedback is collected: - The user interface 160 allows the user to rate (e.g., 0 stars to 5 stars) the success rate of processing, safety features, and / or result quality; - The image, time series signal data, and / or prediction result are labeled with the feedback rating for continuous learning of the model.

[0051] The system 100 can operate on a multi-layered architecture, with the bottom layer being the input layer, consisting of sensors 135-155 and user feedback, which capture multi-modal data. The information collected by these sensors 135-155 includes weight measurements, current readings, temperature changes, visual observations, odor signals, and humidity levels, providing a comprehensive understanding of the cooking environment. User inputs, including cooking preferences, ingredient choices, and desired outcomes, further refine the data collection process. User inputs can be collected through the user interface 160 and / or the mobile device 115.

[0052] The second layer includes the processing device 140, enabling the kitchen appliance 105 to operate as an edge computing device. This processing device 140 serves as the core of edge automation, making real-time decisions on when to initiate or stop various kitchen processes, including ingredient preprocessing and cooking steps. The edge device is preferably equipped with multi-modal inference algorithms and edge reinforcement learning capabilities, enabling personalized recipe recommendations and edge-side recipe personalization adjustments. These algorithms can be optimized through firmware updates for better performance. The edge device can utilize reinforcement learning (RL) and embedding techniques to autonomously manage various cooking tasks, including ingredient preprocessing, ingredient mixing, and cooking temperature control. The reinforcement learning RL algorithm can optimize appliance operations based on environmental feedback, ensuring that the cooking results meet user preferences at every step.

[0053] Meanwhile, sensor data and user input data can be transmitted to the computing backend 110, which can be implemented as a server or service possibly deployed on a computer cloud. After filtering out personally identifiable information (PII), the collected data is aggregated to build a unified model that integrates common sense knowledge and comprehensive databases. This allows online training of a multi-modal inference model that can receive real-time multi-modal data inputs from the target appliance 105 and output personalized food ratings / indexes. The model can then be updated on the mobile device 115 and communicate with the kitchen appliance 105 to adjust the actuators in real-time based on the model output during food processing. Reinforcement learning (RL), embedding techniques, knowledge graphs (KG), large language models (LLM), and large visual models (LVM) can be used to train the model to continuously optimize based on new information. Each of these techniques can be used to train a unified model that integrates common sense knowledge and comprehensive databases.

[0054] Reinforcement Learning (RL): can be used to optimize the decision-making process of the intelligent kitchen system. By employing reinforcement learning RL algorithms, the system can learn from interactions with the environment (i.e., the cooking process) and adjust operations to maximize long-term rewards. For example, reinforcement learning RL can help the system determine the optimal cooking parameters (such as temperature and cooking duration) based on user preferences and ingredient characteristics.

[0055] Embedding techniques: Embedding techniques are dense vector representations of words or entities that capture their semantic relationships. In the smart kitchen system, embedding techniques can be used to represent ingredients, cooking methods, and user preferences in a continuous vector space, enabling the system to better understand the similarities and associations between different cooking elements, leading to more accurate recommendations and personalized services.

[0056] Foundation model: The foundation model refers to a model that captures core knowledge such as cooking principles, ingredient interactions, and flavor characteristics. This model is the core of the unified model proposed in this paper, which can comprehensively understand cooking science, provide decision-making basis, and optimize cooking processes.

[0057] Knowledge graph (KG): A knowledge graph represents structured information about entities and their relationships in a graphical form. In the smart kitchen system, a knowledge graph can capture domain-specific knowledge related to ingredients, recipes, cooking methods, and user preferences. By organizing this information into a graphical structure, the system can efficiently query and retrieve relevant knowledge, improving decision-making capabilities and personalized service levels.

[0058] Large language model (LLM): Large language models such as BERT or GPT based on Transformer are good at understanding natural language text and generating contextually relevant responses. In the smart kitchen system, a large language model LLM can be used to analyze user text inputs such as recipe instructions, ingredient lists, and cooking preferences, extracting valuable information from text data to enable the system to adjust recommendations and operations accordingly.

[0059] Large visual model (LVM): A large visual model is a deep learning model designed specifically to process and understand visual information. In the smart kitchen system, a large visual model LVM can analyze images and videos captured by cameras in the kitchen environment, recognizing ingredients, monitoring cooking progress, and detecting abnormal situations or safety hazards. By integrating a large visual model LVM, the system can visually understand the cooking process, making more reasonable decisions and providing personalized assistance.

[0060] Integrating these technologies into the training process of the unified model enables the smart kitchen system 100 to learn from various data sources such as text, images, and structured knowledge. Through continuous optimization and adjustment, the system can comprehensively understand cooking principles and user preferences, optimizing cooking processes and providing personalized cooking experiences that meet the taste preferences and cooking styles of each user.

[0061] In addition, the application scope of this common sense knowledge base can extend beyond a single appliance. Insights gained from the unified model can be applied to other kitchen appliances such as microwaves and ovens, enabling these appliances to achieve the same level of personalization and intelligence. Federated learning technology can facilitate seamless knowledge transfer between different devices, enabling each appliance to continuously improve and develop based on collective insights and experiences.

[0062] List of reference signs 100 system 105 kitchen appliance 110 computing backend 115 mobile device 120 container 125 drive unit 130 mechanical tool 135 camera 140 processing device 145 weight sensor 150 temperature sensor 152 communication unit 155 VOC sensor 160 user interface 200 flowchart 205 trigger data collection 210 acquire data 215 statistical analysis 220 AI model prediction 225 provide output 230 control processing parameters 235 collect user feedback

Claims

1. A kitchen appliance (105) for processing foodstuff, the kitchen appliance comprising: - a container (120) for the foodstuff; - a drive unit (125) having a mechanical tool (130) to dispose of the foodstuff; - input means (115, 160) for determining a desired state of the foodstuff; - a camera (135) for taking an image of a content within the container (120); - processing means (140) for determining a state of the foodstuff based on the image; and - output means (115, 160) for outputting an indication information of the determined state of the foodstuff.

2. The kitchen appliance (105) of claim 1, wherein, The indication information comprises a notification that the foodstuff has reached the desired state.

3. The kitchen appliance (105) according to claim 1 or 2, wherein The processing means (140) are adapted to control the drive unit (125) such that the foodstuff is processed to reach the desired state.

4. The kitchen appliance (105) according to any one of the preceding claims, wherein, The desired state is determined based on a recipe for preparing a predetermined dish, the dish being based on the foodstuff.

5. The kitchen appliance (105) according to any one of the preceding claims, wherein, The kitchen appliance further comprises a further sensor for detecting a content within the container (120), the processing means (140) being adapted to determine the state of the foodstuff also based on a quantity sensed by the further sensor.

6. The kitchen appliance (105) according to any one of the preceding claims, wherein, The processing means (140) are adapted to execute a machine learning technique to determine the state of the foodstuff.

7. The kitchen appliance (105) of claim 6, wherein, The processing means (140) are adapted to train an artificial neural network using the machine learning technique based on data observed from the foodstuff during a processing.

8. The kitchen appliance (105) according to any one of the preceding claims, wherein, The kitchen appliance further comprises communication means (152) for exchanging data with a remote computing backend (110).

9. A system (100) comprising a kitchen appliance (105) according to claim 8 and a computing backend (110) running a food processing base model, wherein The computing backend (110) is adapted to receive from the kitchen appliance (105): - at least one quantity sensed of a content within the container (120), - an indication information of a processing procedure to be executed for the foodstuff, - an indication information of a user operating the domestic appliance, and The computing backend (110) is adapted to provide, by means of the base model, parameters for the kitchen appliance (105) for executing a processing procedure.

10. The system (100) of claim 9, wherein, The user has an associated food preparation preference.

11. The system (100) according to claim 9 or 10, wherein The user has an associated observable food preparation behavior record.

12. The system (100) according to any one of claims 9 to 11, wherein, The parameters are determined based on a dish to be prepared.

Images