Granted Patent B2

US 12,456,283 · App. 17/986,378 · Granted Oct 28, 2025

Inventors: Yair Adato (Kfar Ben Nun, IL); Ran Achituv (Kefar Sava, IL); Eyal Gutflaish (Beer Sheva, IL); Dvir Yerushalmi (Kfar Saba, IL)

Assignee: BRIA ARTIFICIAL INTELLIGENCE LTD

G06V10/764G06F18/214G06F18/24G06T3/40G06T5/50G06T7/0002G06T11/001G06T19/006G06V10/772G06V10/774G06T2207/20081

Abstract

Claims (100)

1. A non-transitory computer readable medium containing instructions for causing at least one processor to perform operations for identifying visual contents used for training of inference models, the operations comprising:

receiving a specific visual content;

accessing data based on at least one parameter of an inference model, the inference model is a result of training a machine learning algorithm using a plurality of training examples, each training example of the plurality of training examples includes a visual content;

analyzing the data and the specific visual content to determine a likelihood that the specific visual content is included in at least one training example of the plurality of training examples;

generating a digital signal indicative of the likelihood that the specific visual content is included in at least one training example of the plurality of training examples;

calculating a convolution of at least part of the specific visual content to thereby obtain a result value of the calculated convolution of the at least part of the specific visual content;

calculating a mathematical function of the result value of the calculated convolution of the at least part of the specific visual content;

selecting a threshold based on the data;

comparing the threshold with the mathematical function of the result value of the calculated convolution of the at least part of the specific visual content; and

determining the likelihood that the specific visual content is included in at least one training example of the plurality of training examples based on a result of the comparison of the threshold with the mathematical function of the result value of the calculated convolution of the at least part of the specific visual content.

2. The non-transitory computer readable medium of claim 1 , wherein the likelihood is a binary likelihood, and determining the likelihood that the specific visual content is included in at least one training example of the plurality of training examples includes determining whether the specific visual content is included in at least one training example of the plurality of training examples.

3. The non-transitory computer readable medium of claim 1 , wherein the operations further comprise:

using the specific visual content and the at least one parameter of the inference model to obtain an output of the inference model corresponding to the specific visual content; and

basing the data on the output of the inference model corresponding to the specific visual content.

4. The non-transitory computer readable medium of claim 1 , wherein the operations further comprise:

using the specific visual content to generate a plurality of variations of the specific visual content;

for each variation of the plurality of variations, using the variation and the at least one parameter of the inference model to obtain an output of the inference model corresponding to the variation; and

basing the data on the outputs of the inference model corresponding to the plurality of variations.

5. The non-transitory computer readable medium of claim 4 , wherein the operations further comprise:

obtaining a plurality of directions in a mathematical space;

obtaining a specific mathematical object in the mathematical space corresponding to the specific visual content;

for each direction of the plurality of directions, using the specific mathematical object and the direction to determine a mathematical object in the mathematical space corresponding to the specific mathematical object and the direction; and

for each direction of the plurality of directions, generating a visual content corresponding to the mathematical object in the mathematical space corresponding to the specific mathematical object and the direction, thereby generating the plurality of variations of the specific visual content.

6. The non-transitory computer readable medium of claim 5 , wherein the mathematical space is a non-orientable space.

7. The non-transitory computer readable medium of claim 4 , wherein the operations further comprise:

selecting a plurality of objects, no object of the plurality of objects is depicted in the specific visual content; and

for each object in the plurality of objects, analyzing the specific visual content to generate a variation of the specific visual content that includes a depiction of the object, thereby generating the plurality of variations of the specific visual content.

8. The non-transitory computer readable medium of claim 4 , wherein the operations further comprise:

analyzing the specific visual content to detect a plurality of objects depicted in the specific visual content; and

for each object in the plurality of objects, analyzing the specific visual content to generate a variation of the specific visual content that do not include a depiction of the object, thereby generating the plurality of variations of the specific visual content.

9. The non-transitory computer readable medium of claim 1 , wherein the operations further comprise:

using the specific visual content and the at least one parameter of the inference model to obtain a gradient corresponding to the specific visual content; and

basing the data on the gradient corresponding to the specific visual content.

10. The non-transitory computer readable medium of claim 1 , wherein the operations further comprise:

using the specific visual content and the at least one parameter of the inference model to calculate a loss corresponding to the specific visual content and to a loss function associated with the machine learning algorithm; and

basing the data on the loss corresponding to the specific visual content and to the loss function.

11. The non-transitory computer readable medium of claim 1 , wherein the operations further comprise:

analyzing the data and the specific visual content to determine a likelihood that an augmented version of the specific visual content is included in at least one training example of the plurality of training examples; and

generating a second digital signal indicative of the likelihood that an augmented version of the specific visual content is included in at least one training example of the plurality of training examples.

12. The non-transitory computer readable medium of claim 1 , wherein the inference model is a generative model, and wherein for each training example of the plurality of training examples, the visual content included in the training example is a pair of visual contents, an input visual content and a desired output visual content.

13. The non-transitory computer readable medium of claim 1 , wherein the operations further comprise:

analyzing the data and the specific visual content to determine a measure of similarity of the specific visual content to a selected training example of the plurality of training examples; and

generating a second digital signal indicative of the measure of similarity of the specific visual content to the selected training example of the plurality of training examples.

14. The non-transitory computer readable medium of claim 13 , wherein the measure of similarity is indicative of an amount of augmentation.

15. The non-transitory computer readable medium of claim 13 , wherein the second digital signal is indicative of the selected training example of the plurality of training examples.

16. The non-transitory computer readable medium of claim 13 , wherein the selected training example of the plurality of training examples is the most similar training example in the plurality of training examples to the specific visual content.

17. The non-transitory computer readable medium of claim 1 , wherein the operations further comprise:

accessing a data structure associating visual contents with sources to determine that the specific visual content is associated with a specific source; and

including an indication of the specific source in the generated digital signal.

18. A system for identifying visual contents used for training of inference models, the system comprising:

at least one processor configured to perform the operations of:

receiving a specific visual content;

accessing data based on at least one parameter of an inference model, the inference model is a result of training a machine learning algorithm using a plurality of training examples, each training example of the plurality of training examples includes a visual content;

analyzing the data and the specific visual content to determine a likelihood that the specific visual content is included in at least one training example of the plurality of training examples;

generating a digital signal indicative of the likelihood that the specific visual content is included in at least one training example of the plurality of training examples;

calculating a convolution of at least part of the specific visual content to thereby obtain a result value of the calculated convolution of the at least part of the specific visual content;

calculating a mathematical function of the result value of the calculated convolution of the at least part of the specific visual content;

selecting a threshold based on the data;

comparing the threshold with the mathematical function of the result value of the calculated convolution of the at least part of the specific visual content; and

determining the likelihood that the specific visual content is included in at least one training example of the plurality of training examples based on a result of the comparison of the threshold with the mathematical function of the result value of the calculated convolution of the at least part of the specific visual content.

19. A method for identifying visual contents used for training of inference models, the method comprising:

receiving a specific visual content;

accessing data based on at least one parameter of an inference model, the inference model is a result of training a machine learning algorithm using a plurality of training examples, each training example of the plurality of training examples includes a visual content;

analyzing the data and the specific visual content to determine a likelihood that the specific visual content is included in at least one training example of the plurality of training examples;

generating a digital signal indicative of the likelihood that the specific visual content is included in at least one training example of the plurality of training examples;

calculating a convolution of at least part of the specific visual content to thereby obtain a result value of the calculated convolution of the at least part of the specific visual content;

calculating a mathematical function of the result value of the calculated convolution of the at least part of the specific visual content;

selecting a threshold based on the data;

comparing the threshold with the mathematical function of the result value of the calculated convolution of the at least part of the specific visual content; and

determining the likelihood that the specific visual content is included in at least one training example of the plurality of training examples based on a result of the comparison of the threshold with the mathematical function of the result value of the calculated convolution of the at least part of the specific visual content.

20. A non-transitory computer readable medium containing instructions for causing at least one processor to perform operations for identifying visual contents used for training of inference models, the operations comprising:

receiving a specific visual content;

accessing data based on at least one parameter of an inference model, the inference model is a result of training a machine learning algorithm using a plurality of training examples, each training example of the plurality of training examples includes a visual content;

analyzing the data and the specific visual content to determine a likelihood that the specific visual content is included in at least one training example of the plurality of training examples;

generating a digital signal indicative of the likelihood that the specific visual content is included in at least one training example of the plurality of training examples, using the specific visual content to generate a plurality of variations of the specific visual content;

for each variation of the plurality of variations, using the variation and the at least one parameter of the inference model to obtain an output of the inference model corresponding to the variation;

basing the data on the outputs of the inference model corresponding to the plurality of variations;

obtaining a plurality of directions in a mathematical space;

obtaining a specific mathematical object in the mathematical space corresponding to the specific visual content;

for each direction of the plurality of directions, using the specific mathematical object and the direction to determine a mathematical object in the mathematical space corresponding to the specific mathematical object and the direction; and

for each direction of the plurality of directions, generating a visual content corresponding to the mathematical object in the mathematical space corresponding to the specific mathematical object and the direction, thereby generating the plurality of variations of the specific visual content.

21. A non-transitory computer readable medium containing instructions for causing at least one processor to perform operations for identifying visual contents used for training of inference models, the operations comprising:

receiving a specific visual content;

accessing data based on at least one parameter of an inference model, the inference model is a result of training a machine learning algorithm using a plurality of training examples, each training example of the plurality of training examples includes a visual content;

analyzing the data and the specific visual content to determine a likelihood that the specific visual content is included in at least one training example of the plurality of training examples;

generating a digital signal indicative of the likelihood that the specific visual content is included in at least one training example of the plurality of training examples, using the specific visual content to generate a plurality of variations of the specific visual content;

for each variation of the plurality of variations, using the variation and the at least one parameter of the inference model to obtain an output of the inference model corresponding to the variation;

basing the data on the outputs of the inference model corresponding to the plurality of variations;

selecting a plurality of objects, no object of the plurality of objects is depicted in the specific visual content; and

for each object in the plurality of objects, analyzing the specific visual content to generate a variation of the specific visual content that includes a depiction of the object, thereby generating the plurality of variations of the specific visual content.

22. A non-transitory computer readable medium containing instructions for causing at least one processor to perform operations for identifying visual contents used for training of inference models, the operations comprising:

receiving a specific visual content;

accessing data based on at least one parameter of an inference model, the inference model is a result of training a machine learning algorithm using a plurality of training examples, each training example of the plurality of training examples includes a visual content;

analyzing the data and the specific visual content to determine a likelihood that the specific visual content is included in at least one training example of the plurality of training examples;

generating a digital signal indicative of the likelihood that the specific visual content is included in at least one training example of the plurality of training examples;

using the specific visual content to generate a plurality of variations of the specific visual content;

for each variation of the plurality of variations, using the variation and the at least one parameter of the inference model to obtain an output of the inference model corresponding to the variation;

basing the data on the outputs of the inference model corresponding to the plurality of variations;

analyzing the specific visual content to detect a plurality of objects depicted in the specific visual content; and

for each object in the plurality of objects, analyzing the specific visual content to generate a variation of the specific visual content that do not include a depiction of the object, thereby generating the plurality of variations of the specific visual content.

Assignments (1)

Continuity (3)

References Cited (19)