TWI857765B - Frame rate intelligent control method and frame rate intelligent control system - Google Patents

Abstract

A frame rate intelligent control method and a frame rate intelligent control system are provided. The frame rate intelligent control method includes the following steps. A plurality of training data are obtained. Each of the training data includes an electronic device model, a game name, a remaining power and an average frame rate. An inference model is trained according to the training data. Using the inference model, the frame rate of an electronic device playing a game is inferenced to obtain a frame rate inference value. The frame rate of the electronic device is set to be the frame rate inference value. The inference model is optimized according to the operation of the electronic device.

Description

Frame rate intelligent control method and frame rate intelligent control system

The present disclosure relates to a control method and a control system, and in particular to a frame rate intelligent control method and a frame rate intelligent control system.

Players can play games through electronic devices such as computers, smartphones or tablets. These electronic devices consume a lot of power when playing games. Especially in the case of high frame rates, power consumption is accelerated.

However, if the frame rate is reduced to save power, the image quality will be greatly affected. In addition, different electronic devices have very different performances in terms of power and image quality, and it is difficult to use a unified rule to set different electronic devices. Therefore, how to strike a balance between power and frame rate is one of the important directions of the researchers' research in Chile.

This disclosure is about a frame rate intelligent control method and frame rate intelligent control system, which performs training procedures, inference procedures and optimization procedures to perform frame rate intelligent control on electronic devices, so that when the electronic devices play games, they can obtain higher picture quality and ensure safe power consumption.

According to one aspect of the present disclosure, a frame rate intelligent control method is proposed. The frame rate intelligent control method includes the following steps. A plurality of training data are obtained. Each training data includes an electronic device model, a game name, a remaining power and an average frame rate. An inference model is trained based on these training data. A frame rate of an electronic device playing a game is inferred by the inference model to obtain a frame rate inference value. The electronic device sets the frame rate to the frame rate inference value. Based on the operation of the electronic device, the inference model is optimized. The step of optimizing the inference model includes the following steps. Determine whether the frame rate of the electronic device is manually adjusted to a manual frame rate value. Determine whether the electronic device is sufficient to continue playing the game for more than one unit time. If the frame rate is not manually adjusted and the electronic device is not sufficient to continue playing the game for more than that unit time, adjust these training data to retrain the inference model.

According to another aspect of the present disclosure, a frame rate intelligent control system is proposed. The frame rate intelligent control system includes a storage unit, a training unit, an inference model, a frame rate inference unit and an optimization unit. The storage unit is used to store a number of training data. Each training data includes an electronic device model, a game name, a remaining power and an average frame rate of an electronic device. The training unit is used to train an inference model based on these training data. The frame rate inference unit is used to infer a frame rate of an electronic device playing a game through an inference model to obtain an inferred frame rate value. The electronic device sets the frame rate to the frame rate inference value. The optimization unit is used to optimize the inference model according to the operation of the electronic device. The optimization unit includes a frame rate monitoring element and a power monitoring element. The frame rate monitoring element is used to determine whether the frame rate of the electronic device is manually adjusted to a manual frame rate value. The power monitoring element is used to determine whether the electronic device is sufficient to continue playing the game for more than a unit time. The adjustment element is used to adjust these training data to retrain the inference model when the frame rate is not manually adjusted and the electronic device is insufficient to continue playing the game for more than the unit time.

In order to better understand the above and other aspects of this disclosure, the following is a specific example, and the attached drawings are used to explain in detail as follows:

100: Frame rate intelligent control system

110: Storage unit

120: Training unit

130: Frame rate inference unit

140: Optimization unit

141: Frame rate monitoring element

142: Electricity monitoring element

143: Adjustment element

144: New components

800i: Electronics

900: Internet

fri: average frame rate

fri’: frame rate inference value

fri*: Manual frame rate value

gmi,gmi’: game name

MDi: Inference Model

OPi: Operational Data

P1: Training program

P2: Inference procedure

P3: Optimization process

pwi,pwi’: remaining power

pwi*: running power

RDi: Real-time data

S110,S111,S112,S113,S114,S115,S116,S117S120,S130,S131,S132,S133,S134,S135,S136,S137,S138,S140,S141,S142,S143,S144,S145: Steps

TRi: Training data

tyi:Electronic device model

FIG. 1 is a schematic diagram showing a training procedure of an inference model according to a frame rate intelligence control system of an embodiment.

FIG. 2 is a schematic diagram showing the inference process of the inference model according to the frame rate intelligent control system of one embodiment.

FIG. 3 is a schematic diagram showing an optimization process of an inference model according to a frame rate intelligent control system of an embodiment.

FIG. 4 shows a block diagram of a frame rate intelligent control system according to an embodiment.

Figure 5 shows a flow chart of a frame rate intelligent control method according to an embodiment.

Figure 6 shows a detailed flow chart of step S110 according to an embodiment.

Figure 7 illustrates the inference model.

Figure 8 shows a detailed flow chart of step S130 according to an embodiment.

Figure 9 shows a detailed flow chart of step S140 according to an embodiment.

Please refer to Figure 1, which shows a schematic diagram of a training procedure P1 for an inference model MDi according to a frame rate intelligent control system 100 of an embodiment. The frame rate intelligent control system 100 is, for example, a server, a cloud computing center, or an edge computing center. The electronic device 800i is, for example, a laptop, a desktop computer, a smart phone, a tablet computer, a vehicle-mounted device, or a head-mounted display. When the electronic device 800i plays a game, a number of training data TRi can be uploaded to the frame rate intelligent control system 100 via the network 900. The models of these electronic devices 800i can be different. The frame rate intelligent control system 100 can train a number of inference models MDi based on these training data TRi. The inference model MDi is mainly used to infer the appropriate frame rate of each electronic device 800i so that the electronic device 800i can achieve a balance between power and frame rate. The frame rate intelligent control system 100 is trained individually for different electronic device models to obtain the inference model MDi corresponding to the different electronic device models.

Please refer to Figure 2, which shows a schematic diagram of the inference process P2 of the frame rate intelligent control system 100 according to an embodiment of the inference model MDi. When an electronic device 800i is in battery discharge mode (i.e., not connected to the mains for charging) and is playing a game, it can upload a real-time data RDi to the frame rate intelligent control system 100 through the network 900. After the frame rate intelligent control system 100 infers a suitable frame rate inference value fri' through the inference model MDi, the electronic device 800i can set the frame rate according to the frame rate inference value fri', in order to obtain higher picture quality and ensure power consumption.

Please refer to Figure 3, which shows a schematic diagram of an optimization procedure P3 of the frame rate intelligent control system 100 according to an embodiment of the invention for the inference model MDi. When an electronic device 800i plays a game according to the frame rate inference value fri', the electronic device 800i can upload an operation data OPi to the frame rate intelligent control system 100 through the network 900. The frame rate intelligent control system 100 can optimize the inference model MDi according to the operation data OPi.

Through the above training program P1, inference program P2 and optimization program P3, the frame rate of the electronic device 800i can be intelligently controlled so that the electronic device 800i can obtain higher picture quality when playing games and ensure power consumption.

Please refer to FIG. 4, which shows a block diagram of a frame rate intelligent control system 100 according to an embodiment. The frame rate intelligent control system 100 includes a storage unit 110, a training unit 120, the above-mentioned inference model MDi, a frame rate inference unit 130 and an optimization unit 140. The optimization unit 140 includes a frame rate monitoring element 141, a power monitoring element 142, an adjustment element 143 and a new element 144. The storage unit 110 is used to store various data, such as a memory, a hard disk or a cloud storage center. The training unit 120 is used to perform a training procedure P1 (shown in FIG. 1 ) of the model, and is, for example, a neural network chip, a circuit, a circuit board, or a storage device for storing program codes. The frame rate inference unit 130 is used to perform an inference procedure P2 (shown in FIG. 2 ) by inferring the model MDi, and is, for example, a neural network chip, a circuit, a circuit board, or a storage device for storing program codes. The optimization unit 140 is used to perform an optimization procedure P3 (shown in FIG. 3 ) on the inference model MDi, and is, for example, a neural network chip, a circuit, a circuit board, or a storage device for storing program codes. The training unit 120, the frame rate inference unit 130 and the optimization unit 140 respectively perform the above-mentioned training procedure P1, inference procedure P2 and optimization procedure P3, that is, the frame rate of the electronic device 800i can be intelligently controlled, so that the electronic device 800i can obtain higher picture quality and ensure power consumption when playing games. The following is a detailed description of the operation of each component with a flowchart.

Please refer to Figure 5, which shows a flow chart of a frame rate intelligent control method according to an embodiment. In step S110, a plurality of training data TRi are obtained. Each training data TRi includes an electronic device model tyi, a game name gmi, a remaining power pwi and an average frame rate fri. The electronic device model tyi is, for example, a product model, a material number, a graphics processor specification, a central processing unit specification, a memory specification, a battery specification or a combination thereof. The game name gmi is, for example, a game program file name, a game window title, or a task name for running the game. The remaining power pwi is, for example, a percentage of the battery power. The average frame rate fri is, for example, the average value of the frame rate per unit time. The unit time is, for example, 1 hour or half an hour. The electronic device 800i uploads a training data TRi to the frame rate intelligent control system 100 every unit time. For example, when the user plays a game with the electronic device 800i for 3 hours, three training data TRi will be transmitted in the first hour, the second hour and the third hour respectively. The remaining power pwi and the average frame rate fri of these three training data TRi may not be the same.

Please refer to Figure 6, which shows a detailed flow chart of step S110 according to an embodiment. Step S110 includes steps S111 to S117. In step S111, the electronic device 800i determines whether a game is being played. If the electronic device 800i is playing a game, it proceeds to step S112. In step S111, the electronic device 800i can analyze whether the application being executed has a game through the task manager. In this step, as long as there is more than one game being played, it can be determined that a game is being played.

Next, in step S112, the electronic device 800i obtains the electronic device model tyi. The electronic device 800i can obtain the electronic device model tyi through system information.

Then, in step S113, the electronic device 800i obtains the game name gmi. When the electronic device 800i is playing multiple games, the game name gmi is obtained based on the game with the largest load.

Next, in step S114, the electronic device 800i determines whether the electronic device 800i is in the battery discharge mode. If the electronic device 800i is in the battery discharge mode, the process proceeds to step S115.

In step S115, the electronic device 800i determines whether the game has been played for a unit time. The unit time is, for example, 1 hour or half an hour. If the game has been played for a unit time, the process proceeds to step S116.

In step S116, the electronic device 800i calculates the average frame rate fri per unit time. The average frame rate fri is the average value of the unit time calculated from the current time point forward.

Next, in step S117, the electronic device 800i uploads a training data TRi. The training data TRi includes the electronic device model tyi, the game name gmi, the remaining power pwi and the average frame rate fri.

Through the above steps S111~S117, the uploading of a training data TRi is completed. If the electronic device 800i continues to play the same game, another training data TRi will be uploaded again every unit time. The training data TRi uploaded again will still include the electronic device model tyi, game name gmi, remaining power pwi and average frame rate fri. As shown in Figure 4, the above training data TRi will be stored in the storage unit 110 of the frame rate intelligent control system 100.

Next, in step S120 of FIG. 5, the training unit 120 of the frame rate intelligent control system 100 trains the inference model MDi based on the training data TRi. As shown in FIG. 4, the training unit 120 obtains multiple training data TRi from the storage unit 110. These training data TRi may correspond to different electronic device models tyi. The training unit 120 classifies the electronic device models tyi and trains different inference models MDi respectively.

Then, in step S130 of FIG. 5, the frame rate inference unit 130 of the frame rate intelligent control system 100 infers the frame rate of the electronic device 800i playing the game by inference model MDi. Please refer to FIG. 7, which illustrates the inference model MDi. The inference model MDi is used to output the frame rate inference value fri' when receiving the real-time game name gmi' and the remaining power pwi'. Different electronic device models tyi correspond to different inference models MDi. Therefore, even if the game name gmi' and the remaining power pwi' are the same, different frame rate inference values fri' may be obtained under different electronic device models tyi.

Please refer to Figure 8, which shows a detailed flow chart of step S130 according to an embodiment. Step S130 includes steps S131 to S138. In step S131, the electronic device 800i determines whether the electronic device 800i is playing a game. If the electronic device 800i is playing a game, it proceeds to step S132. In step S131, the electronic device 800i can analyze whether the application being executed has a game through the task manager. In this step, as long as there is more than one game being played, it can be determined that the game is being played.

Next, in step S132, the electronic device 800i obtains the real-time game name gmi' and the remaining power pwi'. When the electronic device 800i is playing multiple games, the game name gmi' is obtained based on the game with the largest load.

Then, in step S133, the electronic device 800i determines whether the electronic device 800i is in the battery discharge mode. If the electronic device 800i is in the battery discharge mode, it proceeds to step S134.

Next, in step S134, the electronic device 800i uploads real-time data RDi. The real-time data RDi includes the electronic device model tyi, the real-time game name gmi’ and the real-time remaining power pwi’.

Then, in step S135, the frame rate inference unit 130 of the frame rate intelligent control system 100 inputs the real-time game name gmi' and the real-time remaining power pwi' to the inference model MDi corresponding to the electronic device model tyi to obtain the frame rate inference value fri'. The frame rate inference value fri' is then returned to the electronic device 800i.

Next, in step S136, the electronic device 800i determines whether the frame rate inference value fri' is a null value. If the frame rate inference value fri' is not a null value, the process proceeds to step S137; if the frame rate inference value fri' is a null value, the process proceeds to step S138. When the frame rate inference value fri' is a null value, it indicates that the power of the electronic device 800i is insufficient, and the inference model MDi cannot infer a suitable frame rate inference value fri'.

In step S137, the electronic device 800i automatically sets the frame rate to the frame rate inference value fri'. In this step, the electronic device 800i automatically controls the display without manual adjustment by the user.

In step S138, the electronic device 800i pops up a charging warning message to inform the user that the power cord must be connected as soon as possible.

According to the above steps S131~S138, the frame rate intelligent control system 100 can infer the appropriate frame rate inference value fri' in real time to provide the electronic device 800i with automatic settings.

Next, in step S140 of FIG. 5, the optimization unit 140 of the frame rate intelligent control system 100 optimizes the inference model MDi according to the operation of the electronic device 800i. After the inference model MDi is trained, further adjustments are still required to improve the inference accuracy.

Please refer to FIG. 9, which shows a detailed flow chart of step S140 according to an embodiment. Step S140 includes steps S141 to S145. In step S141, as shown in FIG. 4, the frame rate monitoring element 141 of the optimization unit 140 of the frame rate intelligent control system 100 determines whether the frame rate of the electronic device 800i is manually adjusted to a manual frame rate value fri*. If the frame rate of the electronic device 800i is not manually adjusted to the manual frame rate value fri*, then step S142 is entered; if the frame rate of the electronic device 800i is manually adjusted to the manual frame rate value fri*, then step S144 is entered.

In this step, as shown in FIG. 4, when the frame rate of the electronic device 800i is manually adjusted to the manual frame rate value fri*, the electronic device 800i will immediately upload the manual frame rate value fri* to the frame rate monitoring element 141 of the optimization unit 140. When the frame rate monitoring element 141 does not receive the manual frame rate value fri*, it can be known that the frame rate of the electronic device 800i has not been manually adjusted; when the frame rate monitoring element 141 receives the manual frame rate value fri*, it can be known that the frame rate of the electronic device 800i has been manually adjusted.

In step S142, the power monitoring element 142 of the optimization unit 140 of the frame rate intelligent control system 100 determines whether the electronic device 800i is sufficient to continue playing the game for more than a unit time. If the electronic device 800i is insufficient to continue playing the game for more than a unit time, the process proceeds to step S143.

In step S142, as shown in FIG. 4, the electronic device 800i will periodically upload the execution power pwi* to the power monitoring element 142 of the optimization unit 140. The power monitoring element 142 determines whether the electronic device 800i is sufficient to continue playing the game for more than a unit time based on the execution power pwi*.

Entering step S143 means that the frame rate of the electronic device 800i still uses the frame rate inference value fri' and has not been manually adjusted, but the electronic device 800i is not enough to continue the game for more than a unit time. In step S143, the adjustment element 143 of the optimization unit 140 of the frame rate intelligent control system 100 adjusts the training data TRi to retrain the inference model MDi. For example, the adjustment element 143 reduces the weight of the training data TRi corresponding to the frame rate inference value fri' to reduce the probability of being inferred in the future.

In step S144, the power monitoring element 142 of the optimization unit 140 of the frame rate intelligent control system 100 determines whether the electronic device 800i is sufficient to continue playing the game for more than a unit time. If the electronic device 800i is not sufficient to continue playing the game for more than a unit time, then enter step S145.

In step S144, as shown in FIG. 4, the electronic device 800i will periodically upload the execution power pwi* to the power monitoring element 142 of the optimization unit 140. The power monitoring element 142 determines whether the electronic device 800i is sufficient to continue playing the game for more than a unit time based on the execution power pwi*.

Entering step S145 indicates that the frame rate of the electronic device 800i is manually adjusted to the manual frame rate value fri*, and the electronic device 800i is sufficient to continue the game for more than the unit time. In step S145, the new component 144 of the optimization unit 140 of the frame rate intelligent control system 100 adds training data TRi according to the manual frame rate value fri* to increase the probability of being inferred in the future.

According to the above embodiment, the frame rate intelligent control system 100 performs the above training procedure P1, inference procedure P2 and optimization procedure P3 respectively, and can perform intelligent frame rate control on the electronic device 800i, so that the electronic device 800i can obtain higher picture quality when playing games and ensure power consumption.

In summary, although the present disclosure has been disclosed as above by the embodiments, it is not intended to limit the present disclosure. Those with ordinary knowledge in the technical field to which the present disclosure belongs can make various changes and modifications without departing from the spirit and scope of the present disclosure. Therefore, the protection scope of the present disclosure shall be subject to the scope defined by the attached patent application.

100: Frame rate intelligent control system

110: Storage unit

120: Training unit

130: Frame rate inference unit

140: Optimization unit

141: Frame rate monitoring element

142: Electricity monitoring element

143: Adjustment element

144: New components

800i: Electronics

fri: average frame rate

fri’: frame rate inference value

fri*: Manual frame rate value

gmi,gmi’: game name

MDi: Inference Model

OPi: Operational Data

pwi,pwi’: remaining power

pwi*: running power

RDi: Real-time data

TRi: Training data

tyi:Electronic device model

Claims (13)

A frame rate intelligent control method includes: obtaining a plurality of training data, each of which includes an electronic device model, a game name, a remaining power and an average frame rate; training an inference model based on the training data; inferring a frame rate of an electronic device playing a game by using the inference model to obtain a frame rate inference value, and the electronic device sets the frame rate to the frame rate inference value; and retraining the inference model based on a manual frame rate value. The frame rate intelligent control method as described in claim 1, wherein retraining the inference model according to the manual frame rate value includes: determining whether the frame rate of the electronic device is manually adjusted to a manual frame rate value; determining whether the electronic device is sufficient to continue the game for more than a unit time; if the frame rate is not manually adjusted, and the electronic device is insufficient to continue the game for more than the unit time, adjusting the training data to retrain the inference model; wherein in the step of adjusting the training data, the weight of the training data corresponding to the inference frame rate value is reduced. The frame rate intelligent control method as described in claim 1, wherein the step of retraining the inference model according to the manual frame rate value further includes: If the frame rate is manually adjusted and the electronic device is sufficient to continue the game for more than the unit time, then according to the manual frame rate value, the training data are added to retrain the inference model. The frame rate intelligent control method as described in claim 3, wherein in the step of adding the training data, the weight of the training data corresponding to the manual frame rate value is increased. The frame rate intelligent control method as described in claim 1, wherein the unit time is 1 hour. The frame rate intelligent control method as described in claim 1, wherein the step of obtaining the pen training data is performed in a battery discharge mode. The frame rate intelligent control method as described in claim 1, wherein the step of inferring the frame rate of the electronic device playing the game is performed in a battery discharge mode. The frame rate intelligent control method as described in claim 1, wherein the step of optimizing the inference model is performed in a battery discharge mode. A frame rate intelligent control system includes: a storage unit for storing a plurality of training data, each of which includes an electronic device model, a game name, a remaining power and an average frame rate of an electronic device; a training unit for training an inference model based on the training data; the inference model; a frame rate inference unit for inferring a frame rate of an electronic device playing a game by using the inference model to obtain an inferred frame rate value, and the electronic device sets the frame rate to the frame rate inference value; and an optimization unit for retraining the inference model based on a manual frame rate value. The frame rate intelligent control system as described in claim 9, wherein the optimization unit includes: a frame rate monitoring element for determining whether the frame rate of the electronic device is manually adjusted to the manual frame rate value; a power monitoring element for determining whether the electronic device is sufficient to continue the game for more than a unit time; an adjustment element for adjusting the training data to retrain the inference model when the frame rate is not manually adjusted and the electronic device is insufficient to continue the game for more than the unit time, and the adjustment element is further used to reduce the weight of the training data corresponding to the inference frame rate value. The frame rate intelligent control system as described in claim 9, wherein the optimization unit further includes: An additional component for adding the training data to retrain the inference model according to the manual frame rate value when the frame rate is manually adjusted and the electronic device is sufficient to continue the game for more than the unit time. The frame rate intelligent control system as described in claim 11, wherein the additional component is further used to increase the weight of the training data corresponding to the manual frame rate value. The frame rate intelligent control system as described in claim 9, wherein the unit time is 1 hour.

Images