Cut-paste training augmentation for machine learning models
Methods and systems of training a machine learning model include identifying an object or person related to an action in a first video. The object or person is copied from the first video to a second video to generate a third video. A machine learning model is trained using the first video and the third video.
1 . A computer-implemented method of training a machine learning model, comprising:
identifying an object or a person related to an action in a first video to determine a bias of the machine learning model to associate the action with the object or person;
copying the object or the person from the first video to a second video to generate a third video that includes labels of the object or the person but excludes a label of the action; and
training a machine learning model using the first video and the third video that mitigates the bias of the machine learning model for the action by disassociating the action with the object or the person.
2 . The method of claim 1 , wherein identifying the object or the person identifies a person performing the action and wherein the second video includes an environment that is different from an environment of the first video.
3 . The method of claim 2 , further comprising labeling the third video with a label associated with the action.
4 . The method of claim 2 , wherein copying the object or the person includes copying portions of frames of the first video that show the action and pasting the portions of frames onto respective frames of the second video.
5 . The method of claim 1 , wherein identifying the object or the person identifies an object associated with a bias of the machine learning model.
6 . The method of claim 5 , further comprising labeling the third video with one or more labels that exclude a label associated with the action.
7 . The method of claim 5 , further comprising:
training the machine learning model using the first video; and
detecting the bias of the machine learning model as a false positive associated with an input that includes the object,
wherein performing the identifying and copying is done responsive to detection of the bias of the machine learning model.
8 . The method of claim 1 , wherein the action relates to an abnormality of a patient health condition in a healthcare setting.
9 . The method of claim 8 , further comprising:
processing video of the patient using the machine learning model;
determining an action relating to the patient health condition; and
notifying a medical professional of the patient health condition to assist the medical professional in decision-making for patient management.
10 . The method of claim 9 , further comprising performing a treatment action responsive to the patient health condition, including an instruction to a treatment system to automatically administer a treatment to the patient.
11 . A system for training a machine learning model, comprising:
a hardware processor; and
a memory that stores a computer program which, when executed by the hardware processor, causes the hardware processor to;
identify an object or a person related to an action in a first video to determine a bias of the machine learning model to associate the action with the object or the person;
copy the object or the person from the first video to a second video to generate a third video that includes labels of the object or the person but excludes a label of the action; and
train a machine learning model using the first video and the third video that mitigates the bias of the machine learning model for the action by disassociating the action with the object or the person.
12 . The system of claim 11 , wherein the computer program causes the hardware processor to identify a person performing the action and wherein the second video includes an environment that is different from an environment of the first video.
13 . The system of claim 12 , wherein the computer program further causes the hardware processor to label the third video with a label associated with the action.
14 . The system of claim 12 , wherein the computer program further causes the hardware processor to copy portions of frames of the first video that show the action and pasting the portions of frames onto respective frames of the second video.
15 . The system of claim 11 , wherein the computer program further causes the hardware processor to identify an object associated with a bias of the machine learning model.
16 . The system of claim 15 , wherein the computer program further causes the hardware processor to label the third video with one or more labels that exclude a label associated with the action.
17 . The system of claim 15 , wherein the computer program further causes the hardware processor to:
train the machine learning model using the first video; and
detect the bias of the machine learning model as a false positive associated with an input that includes the object,
wherein the identification and copying are done responsive to detection of the bias of the machine learning model.
18 . The system of claim 11 , wherein the action relates to an abnormality of a patient health condition in a healthcare setting.
19 . The system of claim 18 , wherein the computer program further causes the hardware processor to:
process video of the patient using the machine learning model;
determine an action relating to the patient health condition; and
notify a medical professional of the patient health condition to assist the medical professional in decision-making for patient management.
20 . The system of claim 19 , wherein the computer program further causes the hardware processor to perform a treatment action responsive to the patient health condition, including an instruction to a treatment system to automatically administer a treatment to the patient.