Granted Patent B2

US 12,361,682 · App. 18/008,730 · Granted Jul 15, 2025

Inventors: Josue Cuevas Juarez (Tokyo, JP); Hasan Arslan (Tokyo, JP); Rajasekhar Sanagavarapu (Tokyo, JP)

Assignee: Rakuten Group, Inc.

G06V10/7715G06V10/44G06V20/60G06V10/764

Abstract

Claims (40)

1. An information processing apparatus comprising:

at least one memory configured to store program code; and

at least one processor configured to read the program code and operate as instructed by the program code, the program code including:

acquisition code configured to cause the at least one processor to acquire a red, blue, green (RGB) image including an object;

converting code configured to cause the at least one processor to apply a discrete cosine transform (DCT) to the RGB image to generate image coefficients corresponding to an YCbCr image comprising Luma (Y) elements and Chroma (Cb, Cr) elements; and

prediction code configured to cause the at least one processor to predict various attributes relating to the object by inputting the image coefficient into a learning model,

wherein the learning model is a learning model that is stored in the at least one memory and shared between a plurality of different objects including the object, and wherein the learning model includes:

a plurality of estimation layers that estimate a plurality of attribute values for a plurality of attributes relating to the plurality of different objects, and

an output layer that concatenates and outputs the plurality of attribute values outputted from the plurality of estimation layers.

2. The information processing apparatus according to claim 1 ,

wherein the learning model is composed of a first part and a second part,

the first part receives the image coefficients as an input and outputs a feature vector expressing features of the object,

the second part includes the plurality of estimation layers and the output layer,

the plurality of estimation layers receive the feature vector as an input and output a value indicating an object type of the object and the plurality of attribute values, and

the output layer concatenates and outputs the value indicating the object type of the object and the plurality of attribute values outputted from the plurality of estimation layers.

3. The information processing apparatus according to claim 2 ,

wherein the prediction code is further configured to cause the at least one processor to predict the various attributes from the plurality of attribute values outputted from the second part of the learning model.

4. The information processing apparatus according to claim 2 ,

wherein at least one valid attribute value out of the plurality of attribute values is set in advance in keeping with the value indicated by the object type, and

wherein the prediction code is further configured to cause the at least one processor to acquire, from the plurality of attribute values, the at least one valid attribute value in keeping with a value indicated by the object type, and predict attributes corresponding to the at least one valid attribute value as the various attributes.

5. The information processing apparatus according to claim 1 ,

wherein in a case where the RGB object image includes a plurality of objects, the prediction code is further configured to cause the at least one processor to predict various attributes that relate to each of the plurality of objects.

6. The information processing apparatus according to claim 1 ,

wherein the image coefficients are concatenated data produced by size matching of the Y elements, the Cb elements, and the Cr elements out of the data produced by the discrete cosine transform.

7. The information processing apparatus according to claim 1 ,

wherein the program code further comprises output configured to cause the at least one processor to output the various attributes.

8. An information processing method performed by at least one processor, the method comprising:

acquiring a red, blue, green (RGB) object image including an object;

applying a discrete cosine transform (DCT) to the RGB image to generate image coefficients corresponding to an YCbCr image comprising Luma (Y) elements and Chroma (Cb, Cr) elements; and

predicting various attributes relating to the object by inputting the image coefficients into a learning model,

wherein the learning model is a learning model that is stored in a memory and shared between a plurality of different objects including the object, and wherein the learning model includes:

a plurality of estimation layers that estimate a plurality of attribute values for a plurality of attributes relating to the plurality of different objects; and

an output layer that concatenates and outputs the plurality of attribute values outputted from the plurality of estimation layers.

9. A non-transitory computer readable medium having instructions stored therein, which when executed by a processor, cause the processor to execute a method comprising:

acquiring a red, blue, green (RGB) an object image including an object;

applying a discrete cosine transform (DCT) to the RGB image to generate image coefficients corresponding to an YCbCr image comprising Luma (Y) elements and Chroma (Cb, Cr) elements; and

predicting various attributes relating to the object by inputting a learning model to the object image,

wherein the learning model is a learning model that is stored in memory and shared between a plurality of different objects including the object and includes:

a plurality of estimation layers that estimate a plurality of attribute values for a plurality of attributes relating to the plurality of different objects; and

an output layer that concatenates and outputs the plurality of attribute values outputted from the plurality of estimation layers.

Assignments (1)

Continuity (1)

References Cited (10)