Data collection system, sensor device, data collection device, and data collection method
A data collection system according to the present disclosure includes: a sensor device that collects data; and a server device including a learning model that performs output according to a learning result, corresponding to input, and a data analysis unit that specifies data that is beneficial for or lacking in training of the learning model. The server device transmits, to the sensor device, a request signal for collecting data that is beneficial for or lacking in the training specified by the data analysis unit, or data similar to the data, the sensor device collects data that is beneficial for or lacking in the training, or similar data based on the received request signal, and transmits the collected data to the server device, and the server device retrains the learning model based on the data transmitted from the sensor device.
1 . A data collection system, comprising:
a sensor device configured to collect first data; and
a server device including:
a learning model configured to generate an output based on a learning result corresponding to an input of the learning model; and
circuitry configured to:
determine minority attribute data based on training data of the learning model, wherein the determined minority attribute data has a proportion smaller than a specific threshold in the training data of the learning model;
transmit, to the sensor device, a request signal that indicates the determined minority attribute data, wherein the sensor device is further configured to:
collect second corresponding to the minority attribute data based on the request signal and the first data; and
transmit the collected second data to the server device; and
train the learning model based on the second data.
2 . The data collection system according to claim 1 , wherein the sensor device is further configured to:
select, as the second data, third data that satisfies at least one of a consent of a subject or a consent of a right holder of the third data, or a judgment criterion of a location associated with the third data; and
transmit the selected third data to the server device.
3 . A sensor device, comprising:
circuitry configured to:
collect first data;
receive, from a server device, a request signal that indicates:
minority attribute data of training data of a learning model, and
a required number of pieces of the minority attribute data, wherein the learning model is associated with the server device;
collect, based on the first data and the request signal, second corresponding to the minority attribute data, wherein the second data includes the required number of pieces of the minority attribute data; and
transmit the collected second data to the server device.
4 . The sensor device according to claim 3 , the circuitry is further configured to:
select, as the second data, third data that satisfies at least one of a consent of a subject or a consent of a right holder of the third data, or a judgment criterion of a location associated with the third data; and
transmit the selected third data to the server device.
5 . The sensor device according to claim 3 , further comprising an image sensor configured to acquire image data, wherein the circuitry is further configured to:
change, based on the request signal, at least one of a resolution of the image data, luminance of the image data, an angle of view of the image data, or a frame rate of the image data; and
transmit, based on the change, the image data to the server device.
6 . The sensor device according to claim 5 , wherein the circuitry is further configured to:
edit the acquired image data based on a specific process; and
transmit the edited image data to the server device.
7 . The sensor device according to claim 6 , wherein the circuitry is further configured to execute the specific process related to privacy protection.
8 . The sensor device according to claim 7 , wherein
the acquired image data includes an image of a first person, and
the circuitry is further configured to deform a face of the first person in the image.
9 . The sensor device according to claim 8 , wherein the circuitry is further configured to apply pixelation to the face of the first person.
10 . The sensor device according to claim 9 , wherein
the circuitry is further configured to change the face of the first person to a face of a second person, and
the second person is different from the first person.
11 . The sensor device according to claim 6 , wherein
the image sensor is further configured to acquire the image data that includes an image of a person, and
an attribute of the person corresponds to a minority attribute associated with the minority attribute data.
12 . The sensor device according to claim 10 , wherein the circuitry is further configured to change the face of the first person in the image based on an absence of a consent of the first person.
13 . A data collection device, comprising:
a learning model configured to generate an output based on a learning result corresponding to an input of the learning model; and
circuitry configured to:
determine minority attribute data based on training data of the learning model, wherein the determined minority attribute data has a proportion smaller than a specific threshold in the training data of the learning model;
transmit, to a sensor device, a request signal that indicates the determined minority attribute data;
receive, from the sensor device, data corresponding to the minority attribute data, wherein the reception of the data is based on the transmitted request signal; and
train the learning model based on the received data.
14 . The data collection device according to claim 13 , wherein
the learning model includes a neural network, and
the circuitry is further configured to:
add the received data to the training data; and
execute, based on the training data that includes the received data, a training process on the neural network.
15 . The data collection device according to claim 13 , wherein the sensor device includes an image sensor.
16 . The data collection device according to claim 15 , wherein the circuitry is further configured to receive, from the sensor device, the data including an image of a first person.
17 . The data collection device according to claim 16 , wherein
the circuitry is further configured to change, in the image, a face of the first person to a face of a second person, and
the second person is different from the first person.
18 . The data collection device according to claim 17 , wherein, in a case where the face of the first person is pixelated in the image, the circuitry is further configured to change a pixelated region of the image to the face of the second person.
19 . A data collection method, comprising:
in a data collection system that includes a server device including a learning model configured to generate an output based on a learning result corresponding to an input of the learning model, and a sensor device:
collecting first data by the sensor device;
determining, by the server device, minority attribute data based on training data of the learning model, wherein the determined minority attribute data has a proportion smaller than a specific threshold in the training data of the learning model;
transmitting, the server device, a request signal that indicates the determined minority attribute data, wherein the request signal is transmitted to the sensor device;
collecting, by the sensor device, second data corresponding to the minority attribute data based on the request signal and the first data;
transmitting, by the sensor device, the collected second data to the server device; and
training, by the server device, the learning model based on the second data.