Patent
Atlas literature
Patent
US 12,488,569 B2Patent drawings and their descriptions. Click a drawing to enlarge it.
FIG. 1 illustrates a method of autonomous vehicle train- ing according to various implementations of the present disclosure;
FIG. 2A illustrates a method of data generation and training according to various implementations of the present disclosure;
FIGS. 3A and 3B illustrate example images used in the synthetic dataset generation engine according to various implementations of the present disclosure;
FIGS. 4A, 4B, and 4C illustrate example images used in the synthetic dataset generation engine according to various implementations of the present disclosure;
FIGS. 5A, 5B, and 5C illustrate example images used in the simulation engine with digital twin dataset according to various implementations of the present …
FIG. 6 illustrates a method of the simulation engine with a digital twin dataset according to various implementations of the present disclosure;
FIG. 7 illustrates a method of training and validating a model according to various implementations of the present disclosure;
FIG. 8 illustrates a method of the synthetic dataset gen- eration engine according to various implementations of the present disclosure;
FIG. 9 illustrates a method of training and validating a model according to various implementations of the present disclosure;
FIG. 10 illustrates a synthetic data generation software architecture according to various implementations of the present disclosure;
FIG. 11 illustrates an example systems architecture according to various implementations of the present disclo- sure; and
FIG. 12 illustrates an electronic device according to various implementations of the present disclosure. Corresponding reference characters indicate …
Claims define the patent's legal scope. Independent claims stand alone; dependent claims (nested) narrow them. Click a claim to expand its dependents.
An apparatus for identifying objects, the apparatus comprising: a processor configured to execute instructions stored on a memory; and the memory storing computer-readable instructions that, when executed by the processor, cause the processor to: receive a dataset comprising an image having first real image features; train a neural network to recognize the first real image features in the dataset; perform a synthetic image augmentation to generate synthetic image features corresponding to the first real image features in the image using the neural network, wherein the synthetic image augmentation allows for training of a computer vision function for recog-nizing second real image features, corresponding to the synthetic image features, in a real-world environment; associate, in real-time, the image to a synthetic image corresponding to the image by correlating features in the image with the synthetic image features, wherein the synthetic image features are of the synthetic image; measure a realism gap between the image and the synthetic image corresponding to the image, wherein to measure the realism gap, the instructions further cause the processor to: classify each of the synthetic image features of the synthetic image to obtain a synthetic classifica-tion, and compare the synthetic classification to a classifica-tion of the image, wherein the classification of the image relates to classification of each of the first real image features; determine that the realism gap does not satisfy a threshold; and update the synthetic image augmentation with data that includes one or more additional semantic objects based on the realism gap not satisfying the threshold.
The apparatus of claim 1, wherein the processor is further configured to: generate a virtual airport environment corresponding to a computationally derived position of a vehicle or recorded position from the vehicle, wherein the vehicle includes a vehicle camera; and generate a corresponding synthetic image based on the computationally derived position of the vehicle.
The apparatus of claim 1, wherein the processor is further configured to import pre-labeled data to label an association of the synthetic image features corresponding to the first real image features in the synthetic image augmen-tation.
The apparatus of claim 1, wherein comparing the synthetic classification to the classification of the image comprises determining a number of the synthetic image features that are incorrectly classified.
The apparatus of claim 1, wherein responsive to a determination that the realism gap does not satisfy the threshold, the processor is configured to: generate a second synthetic image corresponding to the image; measure a second realism gap between the image and the 65 second synthetic image corresponding to the image; and B₂ determine whether the second realism gap satisfies the threshold.
The apparatus of claim 1, wherein the one or more additional semantic objects are primary objects which are added to the data that includes the one or more additional semantic objects that are a focus of the image.
A method for identifying objects comprising receiving, by a device, a dataset comprising an image having first real image features; training, by the device, a neural network to recognize the first real image features in the image; performing, by the device, a synthetic image augmenta-tion to generate synthetic image features corresponding to the first real image features in the image using the neural network, wherein the synthetic image augmentation allows for training of a computer vision function for recogniz-ing second real image features, corresponding to the synthetic image features, in a real-world environ-ment; associating, in real-time and by the device, the image to a synthetic image corresponding to the image by cor-relating features in the image with the synthetic image features, wherein the synthetic image features are of the synthetic image; measuring, by the device, a realism gap between the image and the synthetic image corresponding to the image, wherein measuring the realism gap comprises: classifying each of the synthetic image features of the synthetic image to obtain a synthetic classifi-cation, and comparing the synthetic classification to a classifi-cation of the image, wherein the classification of the image relates to classification of each of the first real image features; determining that the realism gap does not satisfy a thresh-old; and updating the synthetic image augmentation with data that includes one or more additional semantic objects based on the realism gap not satisfying the threshold.
The method of claim 9, further comprising: generating a virtual airport environment corresponding to a computationally derived position of a vehicle or recorded position from the vehicle, wherein the vehicle includes a vehicle camera; and generating a corresponding synthetic image based on the recorded position of the vehicle.
The method of claim 9, further comprising importing prelabeled data to label the synthetic image features corre-sponding to the real image features in the synthetic image augmentation.
The method of claim 9, wherein comparing the synthetic classification to the classification of the image comprises: determining a number of the synthetic image features that are incorrectly classified.
The method of claim 9, wherein responsive to a determination that the realism gap does not satisfy the threshold, the method further comprises: generating a second synthetic image corresponding to the image; measuring a second realism gap between the image and the second synthetic image corresponding to the image; and determining whether the second realism gap satisfies the threshold.
A non-transitory computer-readable medium storing a set of instructions, the instructions comprising: one or more instructions that, when executed by one or more processors of a device, cause the device to: receive a dataset comprising an image having first real image features; train a neural network to recognize the first real image features in the image; perform a synthetic image augmentation to generate syn-thetic image features corresponding to the first real image features in the image using the neural network, such that the synthetic image augmentation allows for improved training of a computer vision function for recognizing second real image features, corresponding to the synthetic image features, in a real-world envi-ronment, associate, in real-time, the image to a synthetic image corresponding to the image by correlating features in the image with the synthetic image features, wherein the synthetic image features are of the syn-thetic image; measure a realism gap between the image and the syn-thetic image corresponding to the image, wherein the one or more instructions, that cause the device to measure the realism gap, cause the device to: classify each of the synthetic image features of the synthetic image to obtain a synthetic classifica-tion, and compare the synthetic classification to a classifica-tion of the image, wherein the classification of the image relates to classification of each of the first real image features; determine that the realism gap does not satisfy a thresh-old; and update the synthetic image augmentation with data that includes one or more additional semantic objects based on the realism gap not satisfying the threshold.
The non-transitory computer-readable medium of claim 17, wherein the one or more instructions, when executed by the one or more processors, further cause the device to: generate a virtual airport environment corresponding to a computationally derived position of a vehicle or recorded position from the vehicle, wherein the vehicle includes a vehicle camera; and generate a corresponding synthetic image based on the computationally derived position of the vehicle.
The non-transitory computer-readable medium of claim 17, wherein the one or more instructions, that cause the device to compare the synthetic classification to the classification of the image, cause the device to: determine a number of the synthetic image features that are incorrectly classified.
The non-transitory computer-readable medium of claim 17, wherein the one or more additional semantic objects are primary objects which are added to the data that includes the one or more additional semantic objects that are a focus of the image. ∗ ∗ ∗ ∗ ∗
Patents and literature cited by this patent (applicant and examiner references).
Cited patents · 4
Related documents with shared materials, methods, properties, or citations.
Patent
Atlas literature
Patent
US 12,488,569 B2Patent drawings and their descriptions. Click a drawing to enlarge it.
FIG. 1 illustrates a method of autonomous vehicle train- ing according to various implementations of the present disclosure;
FIG. 2A illustrates a method of data generation and training according to various implementations of the present disclosure;
FIGS. 3A and 3B illustrate example images used in the synthetic dataset generation engine according to various implementations of the present disclosure;
FIGS. 4A, 4B, and 4C illustrate example images used in the synthetic dataset generation engine according to various implementations of the present disclosure;
FIGS. 5A, 5B, and 5C illustrate example images used in the simulation engine with digital twin dataset according to various implementations of the present …
FIG. 6 illustrates a method of the simulation engine with a digital twin dataset according to various implementations of the present disclosure;
FIG. 7 illustrates a method of training and validating a model according to various implementations of the present disclosure;
FIG. 8 illustrates a method of the synthetic dataset gen- eration engine according to various implementations of the present disclosure;
FIG. 9 illustrates a method of training and validating a model according to various implementations of the present disclosure;
FIG. 10 illustrates a synthetic data generation software architecture according to various implementations of the present disclosure;
FIG. 11 illustrates an example systems architecture according to various implementations of the present disclo- sure; and
FIG. 12 illustrates an electronic device according to various implementations of the present disclosure. Corresponding reference characters indicate …
Claims define the patent's legal scope. Independent claims stand alone; dependent claims (nested) narrow them. Click a claim to expand its dependents.
An apparatus for identifying objects, the apparatus comprising: a processor configured to execute instructions stored on a memory; and the memory storing computer-readable instructions that, when executed by the processor, cause the processor to: receive a dataset comprising an image having first real image features; train a neural network to recognize the first real image features in the dataset; perform a synthetic image augmentation to generate synthetic image features corresponding to the first real image features in the image using the neural network, wherein the synthetic image augmentation allows for training of a computer vision function for recog-nizing second real image features, corresponding to the synthetic image features, in a real-world environment; associate, in real-time, the image to a synthetic image corresponding to the image by correlating features in the image with the synthetic image features, wherein the synthetic image features are of the synthetic image; measure a realism gap between the image and the synthetic image corresponding to the image, wherein to measure the realism gap, the instructions further cause the processor to: classify each of the synthetic image features of the synthetic image to obtain a synthetic classifica-tion, and compare the synthetic classification to a classifica-tion of the image, wherein the classification of the image relates to classification of each of the first real image features; determine that the realism gap does not satisfy a threshold; and update the synthetic image augmentation with data that includes one or more additional semantic objects based on the realism gap not satisfying the threshold.
The apparatus of claim 1, wherein the processor is further configured to: generate a virtual airport environment corresponding to a computationally derived position of a vehicle or recorded position from the vehicle, wherein the vehicle includes a vehicle camera; and generate a corresponding synthetic image based on the computationally derived position of the vehicle.
The apparatus of claim 1, wherein the processor is further configured to import pre-labeled data to label an association of the synthetic image features corresponding to the first real image features in the synthetic image augmen-tation.
The apparatus of claim 1, wherein comparing the synthetic classification to the classification of the image comprises determining a number of the synthetic image features that are incorrectly classified.
The apparatus of claim 1, wherein responsive to a determination that the realism gap does not satisfy the threshold, the processor is configured to: generate a second synthetic image corresponding to the image; measure a second realism gap between the image and the 65 second synthetic image corresponding to the image; and B₂ determine whether the second realism gap satisfies the threshold.
The apparatus of claim 1, wherein the one or more additional semantic objects are primary objects which are added to the data that includes the one or more additional semantic objects that are a focus of the image.
A method for identifying objects comprising receiving, by a device, a dataset comprising an image having first real image features; training, by the device, a neural network to recognize the first real image features in the image; performing, by the device, a synthetic image augmenta-tion to generate synthetic image features corresponding to the first real image features in the image using the neural network, wherein the synthetic image augmentation allows for training of a computer vision function for recogniz-ing second real image features, corresponding to the synthetic image features, in a real-world environ-ment; associating, in real-time and by the device, the image to a synthetic image corresponding to the image by cor-relating features in the image with the synthetic image features, wherein the synthetic image features are of the synthetic image; measuring, by the device, a realism gap between the image and the synthetic image corresponding to the image, wherein measuring the realism gap comprises: classifying each of the synthetic image features of the synthetic image to obtain a synthetic classifi-cation, and comparing the synthetic classification to a classifi-cation of the image, wherein the classification of the image relates to classification of each of the first real image features; determining that the realism gap does not satisfy a thresh-old; and updating the synthetic image augmentation with data that includes one or more additional semantic objects based on the realism gap not satisfying the threshold.
The method of claim 9, further comprising: generating a virtual airport environment corresponding to a computationally derived position of a vehicle or recorded position from the vehicle, wherein the vehicle includes a vehicle camera; and generating a corresponding synthetic image based on the recorded position of the vehicle.
The method of claim 9, further comprising importing prelabeled data to label the synthetic image features corre-sponding to the real image features in the synthetic image augmentation.
The method of claim 9, wherein comparing the synthetic classification to the classification of the image comprises: determining a number of the synthetic image features that are incorrectly classified.
The method of claim 9, wherein responsive to a determination that the realism gap does not satisfy the threshold, the method further comprises: generating a second synthetic image corresponding to the image; measuring a second realism gap between the image and the second synthetic image corresponding to the image; and determining whether the second realism gap satisfies the threshold.
A non-transitory computer-readable medium storing a set of instructions, the instructions comprising: one or more instructions that, when executed by one or more processors of a device, cause the device to: receive a dataset comprising an image having first real image features; train a neural network to recognize the first real image features in the image; perform a synthetic image augmentation to generate syn-thetic image features corresponding to the first real image features in the image using the neural network, such that the synthetic image augmentation allows for improved training of a computer vision function for recognizing second real image features, corresponding to the synthetic image features, in a real-world envi-ronment, associate, in real-time, the image to a synthetic image corresponding to the image by correlating features in the image with the synthetic image features, wherein the synthetic image features are of the syn-thetic image; measure a realism gap between the image and the syn-thetic image corresponding to the image, wherein the one or more instructions, that cause the device to measure the realism gap, cause the device to: classify each of the synthetic image features of the synthetic image to obtain a synthetic classifica-tion, and compare the synthetic classification to a classifica-tion of the image, wherein the classification of the image relates to classification of each of the first real image features; determine that the realism gap does not satisfy a thresh-old; and update the synthetic image augmentation with data that includes one or more additional semantic objects based on the realism gap not satisfying the threshold.
The non-transitory computer-readable medium of claim 17, wherein the one or more instructions, when executed by the one or more processors, further cause the device to: generate a virtual airport environment corresponding to a computationally derived position of a vehicle or recorded position from the vehicle, wherein the vehicle includes a vehicle camera; and generate a corresponding synthetic image based on the computationally derived position of the vehicle.
The non-transitory computer-readable medium of claim 17, wherein the one or more instructions, that cause the device to compare the synthetic classification to the classification of the image, cause the device to: determine a number of the synthetic image features that are incorrectly classified.
The non-transitory computer-readable medium of claim 17, wherein the one or more additional semantic objects are primary objects which are added to the data that includes the one or more additional semantic objects that are a focus of the image. ∗ ∗ ∗ ∗ ∗
Patents and literature cited by this patent (applicant and examiner references).
Cited patents · 4
Related documents with shared materials, methods, properties, or citations.
Patent
Atlas literature
Patent
US 12,488,569 B2Patent drawings and their descriptions. Click a drawing to enlarge it.
FIG. 1 illustrates a method of autonomous vehicle train- ing according to various implementations of the present disclosure;
FIG. 2A illustrates a method of data generation and training according to various implementations of the present disclosure;
FIGS. 3A and 3B illustrate example images used in the synthetic dataset generation engine according to various implementations of the present disclosure;
FIGS. 4A, 4B, and 4C illustrate example images used in the synthetic dataset generation engine according to various implementations of the present disclosure;
FIGS. 5A, 5B, and 5C illustrate example images used in the simulation engine with digital twin dataset according to various implementations of the present …
FIG. 6 illustrates a method of the simulation engine with a digital twin dataset according to various implementations of the present disclosure;
FIG. 7 illustrates a method of training and validating a model according to various implementations of the present disclosure;
FIG. 8 illustrates a method of the synthetic dataset gen- eration engine according to various implementations of the present disclosure;
FIG. 9 illustrates a method of training and validating a model according to various implementations of the present disclosure;
FIG. 10 illustrates a synthetic data generation software architecture according to various implementations of the present disclosure;
FIG. 11 illustrates an example systems architecture according to various implementations of the present disclo- sure; and
FIG. 12 illustrates an electronic device according to various implementations of the present disclosure. Corresponding reference characters indicate …
Claims define the patent's legal scope. Independent claims stand alone; dependent claims (nested) narrow them. Click a claim to expand its dependents.
An apparatus for identifying objects, the apparatus comprising: a processor configured to execute instructions stored on a memory; and the memory storing computer-readable instructions that, when executed by the processor, cause the processor to: receive a dataset comprising an image having first real image features; train a neural network to recognize the first real image features in the dataset; perform a synthetic image augmentation to generate synthetic image features corresponding to the first real image features in the image using the neural network, wherein the synthetic image augmentation allows for training of a computer vision function for recog-nizing second real image features, corresponding to the synthetic image features, in a real-world environment; associate, in real-time, the image to a synthetic image corresponding to the image by correlating features in the image with the synthetic image features, wherein the synthetic image features are of the synthetic image; measure a realism gap between the image and the synthetic image corresponding to the image, wherein to measure the realism gap, the instructions further cause the processor to: classify each of the synthetic image features of the synthetic image to obtain a synthetic classifica-tion, and compare the synthetic classification to a classifica-tion of the image, wherein the classification of the image relates to classification of each of the first real image features; determine that the realism gap does not satisfy a threshold; and update the synthetic image augmentation with data that includes one or more additional semantic objects based on the realism gap not satisfying the threshold.
The apparatus of claim 1, wherein the processor is further configured to: generate a virtual airport environment corresponding to a computationally derived position of a vehicle or recorded position from the vehicle, wherein the vehicle includes a vehicle camera; and generate a corresponding synthetic image based on the computationally derived position of the vehicle.
The apparatus of claim 1, wherein the processor is further configured to import pre-labeled data to label an association of the synthetic image features corresponding to the first real image features in the synthetic image augmen-tation.
The apparatus of claim 1, wherein comparing the synthetic classification to the classification of the image comprises determining a number of the synthetic image features that are incorrectly classified.
The apparatus of claim 1, wherein responsive to a determination that the realism gap does not satisfy the threshold, the processor is configured to: generate a second synthetic image corresponding to the image; measure a second realism gap between the image and the 65 second synthetic image corresponding to the image; and B₂ determine whether the second realism gap satisfies the threshold.
The apparatus of claim 1, wherein the one or more additional semantic objects are primary objects which are added to the data that includes the one or more additional semantic objects that are a focus of the image.
A method for identifying objects comprising receiving, by a device, a dataset comprising an image having first real image features; training, by the device, a neural network to recognize the first real image features in the image; performing, by the device, a synthetic image augmenta-tion to generate synthetic image features corresponding to the first real image features in the image using the neural network, wherein the synthetic image augmentation allows for training of a computer vision function for recogniz-ing second real image features, corresponding to the synthetic image features, in a real-world environ-ment; associating, in real-time and by the device, the image to a synthetic image corresponding to the image by cor-relating features in the image with the synthetic image features, wherein the synthetic image features are of the synthetic image; measuring, by the device, a realism gap between the image and the synthetic image corresponding to the image, wherein measuring the realism gap comprises: classifying each of the synthetic image features of the synthetic image to obtain a synthetic classifi-cation, and comparing the synthetic classification to a classifi-cation of the image, wherein the classification of the image relates to classification of each of the first real image features; determining that the realism gap does not satisfy a thresh-old; and updating the synthetic image augmentation with data that includes one or more additional semantic objects based on the realism gap not satisfying the threshold.
The method of claim 9, further comprising: generating a virtual airport environment corresponding to a computationally derived position of a vehicle or recorded position from the vehicle, wherein the vehicle includes a vehicle camera; and generating a corresponding synthetic image based on the recorded position of the vehicle.
The method of claim 9, further comprising importing prelabeled data to label the synthetic image features corre-sponding to the real image features in the synthetic image augmentation.
The method of claim 9, wherein comparing the synthetic classification to the classification of the image comprises: determining a number of the synthetic image features that are incorrectly classified.
The method of claim 9, wherein responsive to a determination that the realism gap does not satisfy the threshold, the method further comprises: generating a second synthetic image corresponding to the image; measuring a second realism gap between the image and the second synthetic image corresponding to the image; and determining whether the second realism gap satisfies the threshold.
A non-transitory computer-readable medium storing a set of instructions, the instructions comprising: one or more instructions that, when executed by one or more processors of a device, cause the device to: receive a dataset comprising an image having first real image features; train a neural network to recognize the first real image features in the image; perform a synthetic image augmentation to generate syn-thetic image features corresponding to the first real image features in the image using the neural network, such that the synthetic image augmentation allows for improved training of a computer vision function for recognizing second real image features, corresponding to the synthetic image features, in a real-world envi-ronment, associate, in real-time, the image to a synthetic image corresponding to the image by correlating features in the image with the synthetic image features, wherein the synthetic image features are of the syn-thetic image; measure a realism gap between the image and the syn-thetic image corresponding to the image, wherein the one or more instructions, that cause the device to measure the realism gap, cause the device to: classify each of the synthetic image features of the synthetic image to obtain a synthetic classifica-tion, and compare the synthetic classification to a classifica-tion of the image, wherein the classification of the image relates to classification of each of the first real image features; determine that the realism gap does not satisfy a thresh-old; and update the synthetic image augmentation with data that includes one or more additional semantic objects based on the realism gap not satisfying the threshold.
The non-transitory computer-readable medium of claim 17, wherein the one or more instructions, when executed by the one or more processors, further cause the device to: generate a virtual airport environment corresponding to a computationally derived position of a vehicle or recorded position from the vehicle, wherein the vehicle includes a vehicle camera; and generate a corresponding synthetic image based on the computationally derived position of the vehicle.
The non-transitory computer-readable medium of claim 17, wherein the one or more instructions, that cause the device to compare the synthetic classification to the classification of the image, cause the device to: determine a number of the synthetic image features that are incorrectly classified.
The non-transitory computer-readable medium of claim 17, wherein the one or more additional semantic objects are primary objects which are added to the data that includes the one or more additional semantic objects that are a focus of the image. ∗ ∗ ∗ ∗ ∗
Patents and literature cited by this patent (applicant and examiner references).
Cited patents · 4
Related documents with shared materials, methods, properties, or citations.
Patent
Atlas literature
Patent
US 12,488,569 B2Patent drawings and their descriptions. Click a drawing to enlarge it.
FIG. 1 illustrates a method of autonomous vehicle train- ing according to various implementations of the present disclosure;
FIG. 2A illustrates a method of data generation and training according to various implementations of the present disclosure;
FIGS. 3A and 3B illustrate example images used in the synthetic dataset generation engine according to various implementations of the present disclosure;
FIGS. 4A, 4B, and 4C illustrate example images used in the synthetic dataset generation engine according to various implementations of the present disclosure;
FIGS. 5A, 5B, and 5C illustrate example images used in the simulation engine with digital twin dataset according to various implementations of the present …
FIG. 6 illustrates a method of the simulation engine with a digital twin dataset according to various implementations of the present disclosure;
FIG. 7 illustrates a method of training and validating a model according to various implementations of the present disclosure;
FIG. 8 illustrates a method of the synthetic dataset gen- eration engine according to various implementations of the present disclosure;
FIG. 9 illustrates a method of training and validating a model according to various implementations of the present disclosure;
FIG. 10 illustrates a synthetic data generation software architecture according to various implementations of the present disclosure;
FIG. 11 illustrates an example systems architecture according to various implementations of the present disclo- sure; and
FIG. 12 illustrates an electronic device according to various implementations of the present disclosure. Corresponding reference characters indicate …
Claims define the patent's legal scope. Independent claims stand alone; dependent claims (nested) narrow them. Click a claim to expand its dependents.
An apparatus for identifying objects, the apparatus comprising: a processor configured to execute instructions stored on a memory; and the memory storing computer-readable instructions that, when executed by the processor, cause the processor to: receive a dataset comprising an image having first real image features; train a neural network to recognize the first real image features in the dataset; perform a synthetic image augmentation to generate synthetic image features corresponding to the first real image features in the image using the neural network, wherein the synthetic image augmentation allows for training of a computer vision function for recog-nizing second real image features, corresponding to the synthetic image features, in a real-world environment; associate, in real-time, the image to a synthetic image corresponding to the image by correlating features in the image with the synthetic image features, wherein the synthetic image features are of the synthetic image; measure a realism gap between the image and the synthetic image corresponding to the image, wherein to measure the realism gap, the instructions further cause the processor to: classify each of the synthetic image features of the synthetic image to obtain a synthetic classifica-tion, and compare the synthetic classification to a classifica-tion of the image, wherein the classification of the image relates to classification of each of the first real image features; determine that the realism gap does not satisfy a threshold; and update the synthetic image augmentation with data that includes one or more additional semantic objects based on the realism gap not satisfying the threshold.
The apparatus of claim 1, wherein the processor is further configured to: generate a virtual airport environment corresponding to a computationally derived position of a vehicle or recorded position from the vehicle, wherein the vehicle includes a vehicle camera; and generate a corresponding synthetic image based on the computationally derived position of the vehicle.
The apparatus of claim 1, wherein the processor is further configured to import pre-labeled data to label an association of the synthetic image features corresponding to the first real image features in the synthetic image augmen-tation.
The apparatus of claim 1, wherein comparing the synthetic classification to the classification of the image comprises determining a number of the synthetic image features that are incorrectly classified.
The apparatus of claim 1, wherein responsive to a determination that the realism gap does not satisfy the threshold, the processor is configured to: generate a second synthetic image corresponding to the image; measure a second realism gap between the image and the 65 second synthetic image corresponding to the image; and B₂ determine whether the second realism gap satisfies the threshold.
The apparatus of claim 1, wherein the one or more additional semantic objects are primary objects which are added to the data that includes the one or more additional semantic objects that are a focus of the image.
A method for identifying objects comprising receiving, by a device, a dataset comprising an image having first real image features; training, by the device, a neural network to recognize the first real image features in the image; performing, by the device, a synthetic image augmenta-tion to generate synthetic image features corresponding to the first real image features in the image using the neural network, wherein the synthetic image augmentation allows for training of a computer vision function for recogniz-ing second real image features, corresponding to the synthetic image features, in a real-world environ-ment; associating, in real-time and by the device, the image to a synthetic image corresponding to the image by cor-relating features in the image with the synthetic image features, wherein the synthetic image features are of the synthetic image; measuring, by the device, a realism gap between the image and the synthetic image corresponding to the image, wherein measuring the realism gap comprises: classifying each of the synthetic image features of the synthetic image to obtain a synthetic classifi-cation, and comparing the synthetic classification to a classifi-cation of the image, wherein the classification of the image relates to classification of each of the first real image features; determining that the realism gap does not satisfy a thresh-old; and updating the synthetic image augmentation with data that includes one or more additional semantic objects based on the realism gap not satisfying the threshold.
The method of claim 9, further comprising: generating a virtual airport environment corresponding to a computationally derived position of a vehicle or recorded position from the vehicle, wherein the vehicle includes a vehicle camera; and generating a corresponding synthetic image based on the recorded position of the vehicle.
The method of claim 9, further comprising importing prelabeled data to label the synthetic image features corre-sponding to the real image features in the synthetic image augmentation.
The method of claim 9, wherein comparing the synthetic classification to the classification of the image comprises: determining a number of the synthetic image features that are incorrectly classified.
The method of claim 9, wherein responsive to a determination that the realism gap does not satisfy the threshold, the method further comprises: generating a second synthetic image corresponding to the image; measuring a second realism gap between the image and the second synthetic image corresponding to the image; and determining whether the second realism gap satisfies the threshold.
A non-transitory computer-readable medium storing a set of instructions, the instructions comprising: one or more instructions that, when executed by one or more processors of a device, cause the device to: receive a dataset comprising an image having first real image features; train a neural network to recognize the first real image features in the image; perform a synthetic image augmentation to generate syn-thetic image features corresponding to the first real image features in the image using the neural network, such that the synthetic image augmentation allows for improved training of a computer vision function for recognizing second real image features, corresponding to the synthetic image features, in a real-world envi-ronment, associate, in real-time, the image to a synthetic image corresponding to the image by correlating features in the image with the synthetic image features, wherein the synthetic image features are of the syn-thetic image; measure a realism gap between the image and the syn-thetic image corresponding to the image, wherein the one or more instructions, that cause the device to measure the realism gap, cause the device to: classify each of the synthetic image features of the synthetic image to obtain a synthetic classifica-tion, and compare the synthetic classification to a classifica-tion of the image, wherein the classification of the image relates to classification of each of the first real image features; determine that the realism gap does not satisfy a thresh-old; and update the synthetic image augmentation with data that includes one or more additional semantic objects based on the realism gap not satisfying the threshold.
The non-transitory computer-readable medium of claim 17, wherein the one or more instructions, when executed by the one or more processors, further cause the device to: generate a virtual airport environment corresponding to a computationally derived position of a vehicle or recorded position from the vehicle, wherein the vehicle includes a vehicle camera; and generate a corresponding synthetic image based on the computationally derived position of the vehicle.
The non-transitory computer-readable medium of claim 17, wherein the one or more instructions, that cause the device to compare the synthetic classification to the classification of the image, cause the device to: determine a number of the synthetic image features that are incorrectly classified.
The non-transitory computer-readable medium of claim 17, wherein the one or more additional semantic objects are primary objects which are added to the data that includes the one or more additional semantic objects that are a focus of the image. ∗ ∗ ∗ ∗ ∗
Patents and literature cited by this patent (applicant and examiner references).
Cited patents · 4
Related documents with shared materials, methods, properties, or citations.
Cited non-patent literature · 5
Cited non-patent literature · 5
Cited non-patent literature · 5
Cited non-patent literature · 5
