A computationally implemented method for training machine learning models in vehicle accident event assessment

Abstract

The present disclosure relates to a computer-implemented method for training a machine learning model in vehicle accident event assessment. The method includes implementing a machine learning model (S100) and preparing training data (S200) by acquiring and automatically learning (S210) sensor data generated from vehicle sensors (2, 3; 4, 5). The method further includes applying data augmentation to the collected data to artificially increase the amount of collected data learned from each sensor, such that an augmented data set is learned for each sensor. For each sensor (2, 3; 4, 5), applying data augmentation (S220) includes transforming the collected data and applying it to multiple different sensor mounting positions and / or multiple different sensor mounting orientations within a specific sensor mounting region (6, 7) (S221) so that an additional data set is acquired, the augmented data set including the collected data and the additional data set. The method further includes training a machine learning model using the augmented data set for each sensor (2, 3; 4, 5) (S300).

Description

[Technical Field]

[0001] The present disclosure relates to an improved computer-implemented method for training machine learning models in vehicle accident event assessment. [Background technology]

[0002] Vehicle accident events occur when riding a bicycle, scooter, motorcycle (MC), etc. due to, for example, skidding, slipping, loss of control of the vehicle, collision with another vehicle or object, etc. Falls involving a moving vehicle are often more abrupt and require more immediate attention than falls occurring to a pedestrian, for example, falling or slipping.

[0003] In particular, MCs can be involved in many collisions, many of which can result in severe injury or even death. Part of the reason for this is that riders on MCs are more vulnerable than drivers in automobiles. Also, MCs lack restraints such as seat belts, as do automobiles. There are also other dynamics that make MC riders far more vulnerable in the event of an accident or collision. MCs with integrated airbag systems are an emerging technology that can reduce the risk of severe injury or death in the event of an accident, especially in the case of a frontal or slightly angled collision, and help the rider resist being thrown from the MC as a result of a collision. This can be lifesaving if used effectively.

[0004] Additionally, the MC driver may wear a vest that can be equipped with one or more airbags or that inflates itself, or a smart helmet with one or more airbags, adapted to provide protection to the MC rider wearing at least one of the vest and helmet if the MC rider is thrown from the MC after a collision. Such a helmet may also include one or more devices that can issue one or more types of warnings. Other protective gear, such as belts, is also contemplated.

[0005] However, a real-time airbag triggering system must detect a crash within milliseconds and then trigger a signal to inflate the airbag.

[0006] Most approaches typically involve observing changes in velocity or acceleration over time to infer a vehicle collision or other type of accident. While this approach works relatively well, it is limited in terms of flexibility, particularly with regard to the variety of situations such as speed, angle, and other ways in which a collision may occur.

[0007] It would therefore be desirable to provide an accurate real-time airbag triggering system for MCs and other two- or three-wheeled vehicles that alleviates the above-mentioned drawbacks. Summary of the Invention

[0008] This is accomplished by a computer-implemented method for training a machine learning model in vehicle accident event assessment, including vehicle accident event detection, for a two-wheeled or three-wheeled vehicle. The method includes implementing a machine learning model and preparing training data. Preparing the training data includes acquiring and automatically learning sensor data generated from sensors located on the vehicle and used to collect data, and applying data augmentation to the collected data to artificially increase the amount of collected data learned from each sensor. In this manner, an augmented data set is learned for each sensor.

[0009] For each sensor, applying data augmentation includes transforming and applying the collected data to a plurality of different sensor mounting positions and / or a plurality of different sensor mounting orientations within a particular sensor mounting region such that an additional data set is obtained. The augmented data set includes the collected data and the additional data set. The method further includes training a machine learning model using the augmented data set for each sensor.

[0010] This means that data augmentation is applied to artificially increase the collected data, training data to be much larger than the original training data, and to add new, artificially generated data that was not there in the first place but is generated to provide a more robust machine learning model.

[0011] For example, the sensors may not be the same and / or may have a slightly different placement, or the collision may have occurred at a slightly higher speed. Machine learning models can be trained using augmented data, making them much more robust to real-world irregularities.

[0012] According to some aspects, applying data augmentation comprises: - applying appropriate rotational extensions to the motion sensors, applying small rotations along certain axes to the sensor data to account for changes in sensor mounting orientation on the vehicle; Applying an appropriate noise extension, such as Gaussian noise, to represent the noisy data from a particular sensor; and - applying a first amplitude scaling data expansion to represent changes in perceived intensity of motion acceleration at different mounting locations of the sensors on the MC that are close to each other.

[0013] Thus, the extensions can be obtained in many different ways, and the extensions described above are suitable, for example, to handle multiple different sensor mounting positions and / or multiple different sensor mounting orientations within a particular sensor mounting area.

[0014] According to some embodiments, at least two sensors are positioned within the sensor mounting region, and the machine learning models of the sensors are trained using the same augmented data set.

[0015] This allows a fully trained, robust backbone machine learning model to be obtained for each sensor mounting area and used for one or more sensors in adjacent locations, where these locations are allowed to vary slightly within the sensor mounting area.

[0016] According to some embodiments, if the sensor data acquired from at least two of the sensors located within the same sensor mounting area indicates the occurrence of an accident within a certain predetermined time period, then at least one safety measure is triggered; otherwise, no safety measure is triggered.

[0017] This minimizes the risk of unintentional triggering of safety measures, such as active safety systems, using multi-sensor devices, which is particularly useful in difficult to detect scenarios that may very well look like a collision or other type of vehicle accident, but are not.

[0018] According to some aspects, applying data augmentation comprises: applying a second amplitude scaling data expansion to represent various speeds of vehicle crash dynamics; - Applying time-compressed / decompressed data expansion to represent the speed at which collision dynamics unfold, and - applying segments of appropriate strength and length and position within the scrambled signal to represent noise or unwanted data in the signal due to errors.

[0019] Thus, the extension can be obtained in many different ways, and the above extension is suitable, for example, when a vehicle accident must be detected by at least two of the sensors within a limited period of time to be considered verified and registered.

[0020] According to some aspects, the sensors include at least one of a motion sensor, a 3D sensor, an accelerometer, a gyroscope, and / or a camera, which means that many types of available and well-known sensors can be used.

[0021] According to some aspects, the additional data set is used to cover possible deployment scenarios for which the initially collected data set lacks data.

[0022] According to some aspects, the additional data sets are used to increase robustness with respect to various changes in vehicle speed, vehicle weight, and signal errors due to temporary or permanent malfunction of sensors during crash scenarios.

[0023] This kind of diverse training data is key to training robust and flexible machine learning models in the future.

[0024] The present disclosure also relates to a computer program, a control unit, and a system associated with the above advantages.

[0025] It is therefore understood that the computer-implemented methods, computer programs, and computer-readable media according to the present disclosure, all as described herein, may be realized in hardware, such as the control unit arrangements and devices, all as described herein, and in the systems described herein. The hardware, such as the control unit arrangements as described herein, and the systems as described herein, are then configured to execute the computer-implemented methods and computer programs, thereby achieving the same advantages and benefits as described for the computer-implemented methods herein. [Brief explanation of the drawings]

[0026] The present disclosure will now be described in more detail with reference to the accompanying drawings. [Figure 1] FIG. 1 shows a schematic representation of a user riding a bike with an integrated sensor. [Figure 2] FIG. 2 shows a schematic representation of the control unit. [Figure 3] FIG. 3 illustrates an exemplary computer program product. [Figure 4A] FIG. 4A shows a flowchart of a method according to the present disclosure. [Figure 4B] FIG. 4B shows a flowchart of a method according to the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0027] Aspects of the present disclosure are more fully described below with reference to the accompanying drawings. However, the different configurations, devices, systems, computer programs, and methods disclosed herein may be embodied in many different forms and should not be construed as limited to the aspects set forth herein. Like numbers in the drawings refer to like elements throughout.

[0028] The terminology used herein is for the purpose of describing aspects of the present disclosure only and is not intended to be limiting of the present invention. As used herein, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.

[0029] Modern data-driven approaches based on machine learning (ML) and deep learning (DL) offer greater flexibility to automatically learn from provided crash data, driving data from various roads, and other stress tests. These methods can also adapt to newer crash scenarios, vehicles, and variables as more and more data becomes available. Essentially, they can be retrained to learn new crash scenarios as newer data is provided.

[0030] Typically, ML-based methods are specific to various aspects of the system on which they are trained: for example, if they are trained on data collected from a particular sensor location on a vehicle, such as in a particular orientation, they cannot be used with good performance at another sensor location, even on the same vehicle with the same sensor type.

[0031] 1, 4A and 4B, the present invention relates to a computer-implemented method for training a machine learning model in vehicle accident event assessment including vehicle accident event detection for a two-wheeled or three-wheeled vehicle 1. The method includes implementing S100 the machine learning model and preparing S200 training data. Preparing S200 the training data includes acquiring and automatically learning S210 sensor data generated from sensors 2, 3; 4, 5 located on the vehicle 1 and used to collect data.

[0032] According to some aspects, vehicle accident event assessment also includes vehicle accident event detection, vehicle accident event characterization, and / or calculation of accident event risk probability.

[0033] Referring to FIG. 1 , there is a subject 8 in the form of a user 8 riding a motorcycle 8, the user 8 wearing a smart protective helmet 9 and a protective vest 10. According to some embodiments, at least one of the helmet 9 and the vest 10 comprises one or more airbags 12, 13 adapted to provide protection to the user 8 wearing at least one of the vest 10 and the helmet 9 when the user 8 is thrown from the MC after a collision. Such a helmet 9 may also comprise one or more devices capable of issuing one or more types of warnings. Other protective gear, such as belts, is also contemplated. One or more airbags 14 may also be integrated into the vehicle 1.

[0034] The present disclosure is applicable to users traveling on or in any type of two- or three-wheeled vehicle.

[0035] The sensor data is typically collected, i.e., obtained, from sensors 2, 3, 4, 5, such as, but not limited to, motion sensors included in, e.g., attached to, the vehicle 1, e.g., 3D sensors such as accelerometers that measure acceleration in three dimensions, and gyroscopes that measure rotational rates in three dimensions. According to some embodiments, the sensors 2, 3; 4, 5 include at least one of a motion sensor, a 3D sensor, an accelerometer, a gyroscope, and / or a camera.

[0036] This means that many types of available and well-known sensors can be used.

[0037] According to some embodiments, as shown in FIG. 2 , the subject 8 wears a means 11 for providing and communicating the location of the subject 8. Such means 11 can comprise any type of suitable positioning system, such as a GPS or a GNSS (Global Navigation Satellite System), and any type of wireless communication system. Such GPS information can also be used to estimate the vehicle's speed, which can provide additional input for collision detection algorithms. For example, an airbag may be triggered only when a speed limit, such as 20 km / h, is exceeded. The means can be comprised in a vest 10 or any suitable type of clothing or wearable device.

[0038] For example, vehicle accident event detection achieved by a vehicle accident event detection algorithm relies on detecting patterns of movement belonging to a vehicle accident, such as a collision, in which the vehicle 1 moves in a sudden and unexpected manner.

[0039] 4A and 4B, preparing training data S200 further includes applying data augmentation S220 to the collected data to artificially increase the amount of collected data acquired from each sensor 2, 3; 4, 5, such that an augmented data set is acquired for each sensor 2, 3; 4, 5. For each sensor 2, 3; 4, 5, applying data augmentation S220 includes transforming and applying the collected data to a plurality of different sensor mounting positions and / or a plurality of different sensor mounting orientations within a particular sensor mounting region 6, 7, such that an additional data set is acquired, the augmented data set including the collected data and the additional data set.

[0040] The configuration parameters used when transforming collected data S221 are, according to some aspects, driven by heterogeneity requirements in the envisioned target deployment scenario.

[0041] The method further includes training S300 a machine learning model using the augmented data set for each sensor 2,3;4,5.

[0042] This means that data augmentation is applied to artificially increase the collected data,training data to be much larger than the original training data,,adding new artificially generated data that was not in the,first place but is generated to provide a more robust machine learning,model that can better predict crash / non-crash scenarios when,realworld heterogeneity occurs.,For the development of high-quality machine learning models, a lot of,related training is important, and collecting vehicle crash data,such as MCs is very expensive.

[0043] According to some aspects, augmented data allows for artificially creating data for scenarios where the originally collected data was not applicable. For example, sensors may not be the same and / or may have a slightly different placement, or a collision may have occurred at a slightly higher velocity. Machine learning models can be trained using the augmented data, making them much more robust to real-world irregularities.

[0044] For example, if the orientation variation in the evolution of a particular sensor location is known or expected to be no more than ±20% along the x-axis, for example, as shown in Figure 1, then rotational expansions are generated only within this range. Furthermore, if a particular sensor location has a known vibration pattern, this can be introduced into the expansion data.

[0045] According to some embodiments, at least two sensors 2, 3; 4, 5 are positioned within a sensor mounting region 6, 7, and machine learning models for the sensors 2, 3; 4, 5 are trained using the same augmented data set. The present disclosure results in a fully trained, robust backbone machine learning model for each sensor mounting region 6, 7. This machine learning model may be used for one or more sensors in adjacent locations, whose locations are allowed to vary slightly within the sensor mounting region 6, 7. When more than two sensors and algorithms are involved, alternative schemes such as majority voting may be used.

[0046] According to some aspects, the utility of such machine learning models is that they are robust to small variations in sensor placement, e.g., small rotations, different types of sensor hardware, and other possibilities in each sensor mounting region 6, 7. This becomes very useful in scenarios where data for algorithm development was collected from a particular sensor and sensor location, but during manufacturing, it is discovered that, for example, due to various constraints and changes in sensor orientation, the sensor cannot be placed in the exact same location, but a nearby location is possible. Even the sensor used during manufacturing may differ from the sensor used to collect data during development. Here, the same carefully trained machine learning model can be deployed on a single sensor or multiple sensors when they are near their intended location, i.e., within the sensor mounting region 6, 7. The machine learning model is robust to small changes in position, orientation, or changes in noise profile from different types of sensors, or small amplitude differences due to changes in the type of sensor deployed, relative to their nearby locations.

[0047] Different corresponding robust backbone machine learning models are required for different sensor mounting areas. This flexibility can be extremely beneficial to vehicle manufacturers during system integration when they must balance various constraints when placing system electronics into a production vehicle.

[0048] The mounting areas 6, 7 are not precisely specified but may have a certain size due to a number of parameters related to the collected data, the type of sensor, etc. For example, a machine learning model generated for a mounting area at a footrest position may not be applicable for a headlamp position to make a crash prediction because the mounting areas are not close enough together.

[0049] According to some embodiments, at least one safety measure is triggered when sensor data acquired from at least two of the sensors 2, 3; 4, 5 positioned within the same sensor mounting area 6, 7 indicates the occurrence of an accident within a certain predetermined period of time, and otherwise no safety measure is triggered.

[0050] Essentially, a collision or other type of vehicle accident needs to be detected by at least two of sensors 2, 3; 4, 5, and within a limited time period, in order to be verified and registered. This minimizes the risk of unintentional triggering of safety measures, such as active safety systems, using multi-sensor devices. This is particularly useful in difficult-to-detect scenarios that look very much like a collision or other type of vehicle accident, but are not.

[0051] According to some aspects, and with new reference to FIGS. 4A and 4B, applying data augmentation S220 may include: applying an appropriate rotational extension to the motion sensor (S222), where a small rotation along a particular axis is applied to the sensor data to account for changes in sensor mounting orientation on the vehicle; applying an appropriate noise extension, such as Gaussian noise, to represent noisy data from a particular sensor S223; - applying a first amplitude scaling data extension S224 to represent variations in perceived intensity of motion acceleration at different mounting locations of the sensors on adjacent MCs.

[0052] According to some embodiments, the above extensions are suitable for handling multiple different sensor mounting locations and / or multiple different sensor mounting orientations within a particular sensor mounting region 6,7.

[0053] According to some further aspects, and with continued reference to FIGS. 4A and 4B, applying data augmentation S220 may include: applying a second amplitude scaling data expansion to represent various speeds of vehicle crash dynamics; Applying time-compressed / decompressed data expansion to represent the speed at which collision dynamics unfold; - applying a segment of appropriate length and position in the scrambled signal with appropriate strength to represent noise or unwanted data in the signal due to errors S227.

[0054] According to some embodiments, the above extension is suitable for cases where a vehicle accident needs to be detected by at least two of sensors 2, 3; 4, 5 and within a limited time period in order to be verified and registered.

[0055] Depending on the requirements of the deployment, two or more of the augmentations may be used, and data augmentations represent carefully selected data transformations that are applied directly to the originally collected data based on the requirements of the deployment scenario.

[0056] The present disclosure enables the creation of machine learning models that are more tolerant to small variations in the position and orientation of vehicle sensors 2, 3; 4, 5. Data augmentation during the development of machine learning models enables backbone machine learning models to be trained first using techniques such as contrastive learning. These machine learning models can then be fine-tuned against relevant training data, such as crash data, abuse data, and road driving data.

[0057] According to some aspects, the additional dataset resulting from data augmentation is used to cover possible deployment scenarios for which there is a lack of data in the initially collected dataset. This lack of data may be due, for example, to cost, time, or technical limitations in conducting those experiments. This kind of diverse training data is key to training robust and flexible machine learning models in the future.

[0058] According to some aspects, the additional data set resulting from data augmentation may also be used to increase robustness with respect to various changes in vehicle speed, vehicle weight, and signal errors due to temporary or permanent malfunction of sensors during crash scenarios.

[0059] FIG. 2 shows schematically the components of a control unit 200 according to one embodiment with respect to some functional units.

[0060] Processing circuitry 210 may be provided using one or more combinations of suitable central processing units (CPUs), multiprocessors, microcontrollers, digital signal processors (DSPs), dedicated hardware accelerators, etc. capable of executing software instructions stored, for example, on a computer program product in the form of storage medium 230. Processing circuitry 210 may further be provided as at least one application specific integrated circuit (ASIC) or field programmable gate array (FPGA).

[0061] In particular, processing circuitry 210 is configured to cause control unit 200 to perform a set of operations or steps. These operations or steps were described above in connection with various radar transceivers and methods. For example, storage medium 230 may store a set of operations, and processing circuitry 210 may be configured to retrieve the set of operations from storage medium 230 and cause control unit 200 to perform the set of operations. The set of operations may be provided as a set of executable instructions. Thus, processing circuitry 210 is configured to perform the methods and operations disclosed herein.

[0062] The storage medium 230 may also comprise persistent storage, which may be, for example, any one or combination of magnetic memory, optical memory, solid-state memory, or even remotely mounted memory.

[0063] The control unit 200 may further comprise a communication interface 220 for communicating with at least one other unit. The communication interface 220 may therefore comprise one or more transmitters and receivers with analog and digital components and an appropriate number of ports for wired or wireless communication.

[0064] Processing circuitry 210 is adapted to control the general operation of control unit 200, for example, by sending data and control signals to external units and storage medium 230, by receiving data and reports from external units, and by retrieving data and instructions from storage medium 230. Other components and associated functions of control unit 200 have been omitted so as not to obscure the concepts presented herein.

[0065] FIG. 3 illustrates a computer program product 310 including computer-executable instructions 320 disposed on a computer-readable medium 330 for performing any of the methods disclosed herein.

[0066] According to some aspects, the present disclosure also relates to a device for vehicle accident event assessment, a device for communicating information including information from a vehicle accident event assessment, and a device for enabling a computer-implemented method for vehicle accident event assessment, wherein the computer-implemented method for fall assessment is as described herein. The device comprises a machine learning model, a sensor that collects data from a user, means for acquiring data, means for processing the data, and means for communicating information. The device comprises a computer program and / or a computer-readable medium described herein, and is for applying the computer-implemented method for vehicle accident event assessment and communicating information, wherein the information includes information from the vehicle accident event assessment described herein.

[0067] According to some aspects, a device for vehicle accident event assessment according to the present disclosure as described herein is disclosed, the device being provided on a vehicle 1 and comprising sensors 2, 3, 6, 7 comprising accelerometers and / or gyroscopes.

[0068] According to some aspects, the present disclosure also relates to a system for vehicle accident event evaluation, a system for communication of information including information from vehicle accident event evaluation, and a system for enabling a computer-implemented method for vehicle accident event evaluation. The computer-implemented method for vehicle accident event evaluation is described herein.

[0069] The system comprises a machine learning model, sensors 2, 3, 4, 5, means for collecting data from the sensors 2, 3, 4, 5, means for obtaining data, means for processing data 200, and means for communicating information. The system comprises a computer program product 310 according to the above.

[0070] The information includes information from a vehicle accident event evaluation as described herein, and / or the computer readable medium 330 comprises a computer program 310 product including computer readable instructions 320 for applying a computer-implemented method for vehicle accident event evaluation and for communicating information, the information including information from a vehicle accident event evaluation as described herein.

[0071] According to some aspects, a system for vehicle accident event evaluation and for communication of information is disclosed, wherein the information includes information from the vehicle accident event evaluation according to the present disclosure as disclosed herein. The system includes one or more devices for vehicle accident event evaluation and for communication of information, wherein the information includes information from the vehicle accident event evaluation as described herein.

[0072] The present disclosure relates to a system for training machine learning models in vehicle accident event assessment, for communicating information, for enabling implementation of the machine learning models as described herein, and for enabling preparation of training data as described herein. The vehicle accident event assessment includes vehicle accident event detection, and the system includes a machine learning model, sensors for collecting data, means for acquiring data, means for processing data, and means for communicating information. The system utilizes a computer program 310 as described herein and / or a computer-readable medium 330 as described herein.

Claims

1. 1. A computer-implemented method for training a machine learning model in vehicle accident event assessment including vehicle accident event detection for a two-wheeled or three-wheeled vehicle (1), the method comprising: Implementing the machine learning model (S100); preparing training data (S200); Preparing training data (S200) Acquiring and automatically learning (S210) sensor data generated from sensors (2, 3; 4, 5) located on the vehicle (1) and used to collect data; applying data augmentation (S220) to the collected data to artificially increase the amount of collected data acquired from each sensor (2, 3; 4, 5) so that an augmented data set is acquired for each sensor (2, 3; 4, 5); Applying data augmentation (S220) for each sensor (2, 3; 4, 5): and applying (S221) the collected data to a plurality of different sensor mounting positions and / or a plurality of different sensor mounting orientations within a particular sensor mounting region (6, 7) so that a further data set is acquired, the extended data set comprising the collected data and the further data set, the method comprising: The computer-implemented method further comprises training (S300) the machine learning model using the augmented data set for each sensor (2, 3; 4, 5).

2. Applying data augmentation (S220) - applying appropriate rotational extensions for the motion sensors (S222), whereby small orientations along certain axes are applied to the sensor data to account for changes in sensor mounting orientation on the vehicle; - Applying an appropriate noise extension, such as Gaussian noise, to represent the noisy data from a particular sensor (S223), and - applying a first amplitude scaling data extension (S224) to represent changes in perceived intensity of motion acceleration at different mounting locations of the sensors on MCs that are close to each other.

3. 3. The computer-implemented method of claim 1, wherein at least two sensors (2, 3; 4, 5) are positioned within a sensor mounting area (6, 7), and the machine learning models for the sensors (2, 3; 4, 5) are trained using the same augmented dataset.

4. 4. The computer-implemented method of claim 3, wherein at least one safety measure is triggered when the sensor data acquired from at least two of the sensors (2, 3; 4, 5) located within the same sensor mounting area (6, 7) indicate the occurrence of an accident within a certain predetermined period of time, and no safety measure is triggered otherwise.

5. Applying data augmentation (S220) - Applying a second amplitude scaling data expansion to represent the various speeds of vehicle crash dynamics (S225); - Applying time compression / decompression data expansion to represent the speed at which collision dynamics unfold (S226), and - applying a segment of appropriate length and position in the scrambled signal with appropriate strength to represent noise or unwanted data in the signal due to errors (S227).

6. The computer-implemented method of any one of claims 1 to 5, wherein the sensors (2, 3; 4, 5) comprise at least one of a motion sensor, a 3D sensor, an accelerometer, a gyroscope, and / or a camera.

7. The computer-implemented method of any one of claims 1 to 6, wherein the further dataset is used to cover possible deployment scenarios for which there is insufficient data in the initially collected dataset.

8. 8. The computer-implemented method of claim 1, wherein the further data set is used to increase robustness with respect to various changes in vehicle speed, vehicle weight, and signal errors due to temporary or permanent malfunctions of sensors during a crash scenario.

9. A computer program (310) comprising computer-readable instructions (320) for applying the computer-implemented method and / or training data preparation according to any one of claims 1 to 8, and / or a computer-readable medium (330) comprising said computer program (310).

10. A computer program (310) comprising computer readable instructions (320) for a machine learning model according to any one of claims 1 to 8, and / or a computer readable medium (330) comprising said computer program (310).

11. A control unit (200) for training a machine learning model, the control unit (200) being adapted to control at least the implementation of a machine learning model according to any one of claims 1 to 8 and / or the preparation of training data according to any one of claims 1 to 8.

12. A system for training a machine learning model in vehicle accident event assessment, for communicating information, for enabling the implementation of a machine learning model according to any one of claims 1 to 8, and for enabling the preparation of training data according to any one of claims 1 to 8, comprising: The vehicle accident event assessment includes vehicle accident event detection; the system includes a machine learning model, a sensor that collects data, a means for acquiring the data, a means for processing the data, and a means for communicating information; A system, wherein the system utilizes a computer program (310) according to claim 9 or 10 and / or a computer readable medium (330) according to claim 9 or 10.

Images