PROCESSING METHOD, PROCESSING SYSTEM, AND PROCESSING PROGRAM
A processing system ( 100 ) is a processing system that executes inference process in an edge device ( 30 ) and a server device ( 20 ). The edge device ( 30 ) includes an inference unit ( 31 ) that executes inference related to a first task on inference target data by using a DNN1; and a determination unit ( 32 ) that transmits an intermediate output value of the DNN1 used to execute the inference related to the first task to the server device ( 20 ) so that the server device ( 20 ) executes a second task which is different from the first task and has a higher operation amount than the first task.
1 . A processing method of executing an inference process, the method comprising:
executing inference according to a first task on inference target data by using a first model; and
transmitting an intermediate output value of the first model used in the executing the inference to an application configured to execute a second task, wherein the second task is distinct from the first task, and the second task has a higher operation amount than the first task.
2 . The processing method according to claim 1 , wherein the application is further configured to execute inference according to the second task using a second model by using the intermediate output value as an input.
3 . The processing method according to claim 2 ,
wherein the first model includes an extraction layer extracting a feature amount which is the intermediate output value from the inference target data and a first processing layer executing the inference of the first task based on the feature amount, and
wherein the second model executes the inference of the second task by using the feature amount as an input.
4 . The processing method according to claim 1 ,
wherein the first task includes:
classifying congestion from inference target image data of the first task, and
outputting a classification result, and
wherein the second task includes:
detecting a subject depicted in the inference target image data of the first task, and
outputting a detection result.
5 . A processing system for executing an inference process, the system comprises a processor configured to execute operations comprising:
executing inference according to a first task on inference target data by using a first model; and
transmitting an intermediate output value of the first model used in the executing the inference to an application configured to execute a second task, wherein the second task is distinct from the first task, and the second task has a higher operation amount than the first task.
6 . A computer-readable non-transitory recording medium storing a computer-executable program instructions that when executed by a processor cause a computer system to execute operations comprising:
executing inference according to a first task on inference target data by using a first model; and
transmitting an intermediate output value of the first model used in the executing the inference to an application configured to execute a second task, wherein the second task is distinct from the first task, and the second task has a higher operation amount than the first task.
7 . The processing method according to claim 1 , wherein the executing inference is performed by an edge device, the device connects to a cloud over a network, the cloud includes a server, the server is distinct from the edge device, and the server performs the second task.
8 . The processing method according to claim 1 , wherein the first model represents a lightweight model using a second deep neural network, and the second model represents a non-lightweight model.
9 . The processing method according to claim 1 ,
wherein the first task includes:
determining a presence of at least a person in a restricted area based on inferencing input image data of the restricted area, and
the second task includes, when the first task detects the presence of a person, determining a number of persons in the restricted area and tracking the person as a target person.
10 . The processing system according to claim 5 , wherein the application is further configured to execute inference according to the second task using a second model by using the intermediate output value as an input.
11 . The processing system according to claim 10 ,
wherein the first model includes an extraction layer extracting a feature amount which is the intermediate output value from the inference target data and a first processing layer executing the inference of the first task based on the feature amount, and
wherein the second model executes the inference of the second task by using the feature amount as an input.
12 . The processing system according to claim 5 ,
wherein the first task includes:
classifying congestion from inference target image data of the first task, and
outputting a classification result, and
wherein the second task includes:
detecting a subject depicted in the inference target image data of the first task, and
outputting a detection result.
13 . The processing system according to claim 5 ,
wherein the executing inference is performed by an edge device, the device connects to a cloud over a network, the cloud includes a server, the server is distinct from the edge device, and the server performs the second task.
14 . The processing system according to claim 5 , wherein the first model represents a lightweight model using a second deep neural network, and the second model represents a non-lightweight model.
15 . The processing system according to claim 5 ,
wherein the first task includes:
determining a presence of at least a person in a restricted area based on inferencing input image data of the restricted area, and
the second task includes, when the first task detects the presence of a person, determining a number of persons in the restricted area and tracking the person as a target person.
16 . The computer-readable non-transitory recording medium according to claim 6 , wherein the application is further configured to execute inference according to the second task using a second model by using the intermediate output value as an input,
wherein the first model includes an extraction layer extracting a feature amount which is the intermediate output value from the inference target data and a first processing layer executing the inference of the first task based on the feature amount, and
wherein the second model executes the inference of the second task by using the feature amount as an input.
17 . The computer-readable non-transitory recording medium according to claim 6 , wherein the first task includes:
classifying congestion from inference target image data of the first task, and
outputting a classification result, and
wherein the second task includes:
detecting a subject depicted in the inference target image data of the first task, and
outputting a detection result.
18 . The computer-readable non-transitory recording medium according to claim 6 , wherein the executing inference is performed by an edge device, the device connects to a cloud over a network, the cloud includes a server, the server is distinct from the edge device, and the server performs the second task.
19 . The computer-readable non-transitory recording medium according to claim 6 , wherein the first model represents a lightweight model using a second deep neural network, and the second model represents a non-lightweight model.
20 . The computer-readable non-transitory recording medium according to claim 6 , wherein the first task includes:
determining a presence of at least a person in a restricted area based on inferencing input image data of the restricted area, and
the second task includes, when the first task detects the presence of a person,
determining a number of persons in the restricted area and tracking the person as a target person.