Training system and data collection device
Provided is a training system that performs training of a machine learning model that generates an impressing content or estimates a camera operation for capturing an impressing content. The training system includes a data collection device that collects data, and a training device that performs training of a machine learning model by using the data collected by the data collection device, in which the training device performs re-training of the machine learning model by using learning data that affects the training of the machine learning model to a predetermined degree or more, insufficient learning data, or data similar thereto, collected on the basis of a result of analyzing learning data that affects the training of the machine learning model.
1 . A training system, comprising:
a data collection device configured to collect learning data; and
a training device configured to:
execute a training operation of a machine learning model, based on the collected learning data;
analyse the collected learning data that affects the training operation of the machine learning model; and
execute a re-training operation of the machine learning model, based on specific learning data, wherein
the specific learning data includes one of
learning data that affects the training operation of the machine learning model to more than a specific degree, or
insufficient learning data, and
the specific learning data is collected based on a result of the analysis of the collected learning data that affects the training operation of the machine learning model.
2 . The training system according to claim 1 , wherein the machine learning model:
generates impressing content, or
estimates a camera operation for capture of the impressing content.
3 . The training system according to claim 1 , wherein the analysis is based on one of
an explainable artificial intelligence (XAI),
a confidence score calculation,
an influence function, or
a Bayesian deep neural network (DNN).
4 . The training system according to claim 1 , wherein
the training device is further configured to
transmit, to the data collection device, a request signal to request for the one of
the learning data that affects the training operation of the machine learning model to more than the specific degree, or
the insufficient learning data,
the data collection device is further configured to transmit, based on the request signal, collected data to the training device, and
the training device is further configured to execute the re-training operation of the machine learning model based on the collected data transmitted from the data collection device.
5 . The training system according to claim 1 , wherein
the data collection device is one of a camera or an imager,
the data collection device is configured to:
capture an image, and
transmit, to the training device, image data of the image captured by change in one of a resolution, a frame rate, a luminance, a color, an angle of view, a viewpoint position, or a line-of-sight direction of the one of the camera or the imager, wherein
the one of the resolution, the frame rate, the luminance, the color, the angle of view, the viewpoint position, or the line-of-sight direction is changed based on a degree of influence on the training operation of the machine learning model.
6 . The training system according to claim 1 , wherein
the training device is further configured to transmit a request signal to request for the specific learning data of the machine learning model to the data collection device,
the data collection device is further configured to transmit, to the training device, the one of the learning data that affects the training operation of the machine learning model to more than the specific degree, or the insufficient learning data, and
the training device is further configured to execute the re-training operation the machine learning model based on the transmitted the one of the learning data that affects the training operation of the machine learning model to more than the specific degree, or the insufficient learning data.
7 . The training system according to claim 6 , wherein the training device is further configured to transmit, based on the request signal, information necessary for the analysis to the data collection device.
8 . A data collection device, comprising:
a processor configured to:
receive, from a training device, a first request signal, wherein
the first request signal requests for learning data of a machine learning model, and
the training device executes a training operation of the machine learning model;
analyse the learning data that affects the training operation of the machine learning model;
collect specific learning data, based on a result of the analysis of the learning data that affects the training operation of the machine learning model, wherein
the specific learning data includes one of
learning data that affects the training operation of the machine learning model to more than a specific degree, or
insufficient learning data; and
control transmission of the collected specific learning data to the training device, wherein the training device executes a re-training operation of the machine learning model based on the specific learning data.
9 . The data collection device according to claim 8 , wherein the processor is further configured to:
receive, from the training device, a second request signal, wherein the second request signal requests for the specific learning data; and
collect the specific learning data based on the received second request signal.
10 . The data collection device according to claim 8 , wherein
the processor is further configured to collect image data captured by change in one of a resolution, a frame rate, a luminance, a color, an angle of view, a viewpoint position, or a line-of-sight direction of one of a camera or an imager that captures an image, and
the image is captured based on a degree of influence on the training operation of the machine learning model.