Information processing apparatus, information processing method, and storage medium
Analysis of data for each category is performed with high accuracy. An information processing apparatus includes one or more memories storing instructions and one or more processors that are in communication with the one or more memories. When executing the instructions, the one or more processors cooperate with the one or more memories to perform operations that comprise identifying a category of input data using a category identification model, and selecting, based on goodness-of-fit values of data belonging to an N-th category with respect to analysis models for first to M-th categories for analysis of data in the first to M-th categories, an analysis model to be applied to the input data identified by the identifying as being in the N-th category from the analysis models for the first to M-th categories.
1 . An information processing apparatus comprising:
one or more memories storing instructions; and
one or more processors that are in communication with the one or more memories and that, when executing the instructions, cooperate with the one or more memories to perform operations that comprise:
training an identification model configured to identify a category of an object in image data, among an identification model configured to identify a category of an object in image data and a detection model for the category configured to detect a specific part of the object after the category has been identified by the identification model;
calculating a goodness-of-fit value, based on a difference between (i) a region of the specific part of an object in a first category detected by inputting image data including the specific part of the object in the first category to a detection model configured to detect a specific part of an object in a second category, and (ii) ground truth information indicating a region in the image data where the specific part of the object in the first category is present, wherein the information processing apparatus does not include a detection model configured to detect the specific part of an object in the first category; and
wherein, in a case where the goodness-of-fit value is equal to or greater than a threshold, training the identification model using supervisory data obtained by labeling the image data including the specific part of the object in the first category with a label of an object in the second category.
2 . The information processing apparatus according to claim 1 , wherein the operations further comprise identifying a category of an image.
3 . The information processing apparatus according to claim 1 , wherein the operations further comprise identifying a category of a document.
4 . The information processing apparatus according to claim 1 , wherein the second category is a category of an analysis model.
5 . The information processing apparatus according to claim 4 , wherein, in a case where identification performance of the identification model with respect to data belonging to the second category decreases as a result of the training, the training of the identification model uses supervisory data labeling data belonging to the first category with the first category to train the identification model.
6 . The information processing apparatus according to claim 1 , wherein the first category is different from a category of an analysis model.
7 . The information processing apparatus according to claim 1 , wherein the identification model is a neural network.
8 . An information processing apparatus comprising:
one or more memories storing instructions; and
one or more processors that are in communication with the one or more memories and that, when executing the instructions, cooperate with the one or more memories to perform operations that comprise:
identifying, using an identification model configured to identify a category of an object in image data, a category of input image data; and
performing control to apply, to the input image data, a detection model configured to detect a specific part of an object in the category identified by the identifying, wherein the identification model has been trained using supervisory data obtained by labeling image data including a specific part of an object in a first category with a label of an object in a second category in a case where a goodness-of-fit value, which is based on a difference between (i) a region of the specific part detected by inputting the image data including the specific part of the object in the first category to a detection model configured to detect the specific part of an object in the second category and (ii) ground truth information indicating a region in the image data where the specific part of the object in the first category is present, is equal to or greater than a threshold.
9 . An information processing method comprising:
training an identification model configured to identify a category of an object in image data, among an identification model configured to identify a category of an object in image data and a detection model for the category configured to detect a specific part of the object after the category has been identified by the identification model;
calculating a goodness-of-fit value, based on a difference between (i) a region of the specific part of an object in a first category detected by inputting image data including the specific part of the object in the first category to a detection model configured to detect a specific part of an object in a second category, and (ii) ground truth information indicating a region in the image data where the specific part of the object in the first category is present, wherein the information processing apparatus does not include a detection model configured to detect the specific part of an object in the first category; and
wherein, in a case where the goodness-of-fit value is equal to or greater than a threshold, training the identification model using supervisory data obtained by labeling the image data including the specific part of the object in the first category with a label of an object in the second category.
10 . A non-transitory computer-readable medium storing computer-executable instructions for causing a computer to perform the method according to claim 9 .
11 . An information processing method comprising:
identifying, using an identification model configured to identify a category of an object in image data, a category of input image data; and
performing control to apply, to the input image data, a detection model configured to detect a specific part of an object in the category identified by the identifying, wherein the identification model has been trained using supervisory data obtained by labeling image data including a specific part of an object in a first category with a label of an object in a second category in a case where a goodness-of-fit value, which is based on a difference between (i) a region of the specific part detected by inputting the image data including the specific part of the object in the first category to a detection model configured to detect the specific part of an object in the second category and (ii) ground truth information indicating a region in the image data where the specific part of the object in the first category is present, is equal to or greater than a threshold.
12 . A non-transitory computer-readable medium storing computer-executable instructions for causing a computer to perform the method according to claim 11 .