Method and system for sketch based search
The present teaching relates to a method, system, and programming for searching content. A sketch of an object is obtained from a user. The sketch is processed by a neural network to generate an image corresponding to the sketch of the object. One or more features are extracted from the image, and one or more images previously stored are identified that have features similar to the one or more features. The one or more images are provided to the user.
1. A method, implemented on a machine having at least one processor, storage, and a communication platform capable of connecting to a network for searching content, the method comprising:
sharing, via a graphical user interface within a social group, a sketch of an object generated by a first user included in the social group;
receiving, via the graphical user interface, a modification to the sketch from at least a second user included in the social group;
encoding the modified sketch to generate a dimensional code reflecting dimensions of the sketch;
processing the dimensional code by a plurality of decoders to generate an image corresponding to the modified sketch of the object;
concatenating the modified sketch and the image to generate a concatenated image;
encoding the concatenated image;
determining a numeral value indicative of authenticity of the image by classifying the encoded concatenated image;
extracting, based on the authenticity, one or more features from the image;
identifying one or more images previously stored that have features similar to the one or more features; and
providing, on a webpage, the one or more images to the social group.
2. The method of claim 1 , wherein the sketch of the object is drawn in the graphical user interface by the first user.
3. The method of claim 2 , further comprising:
receiving, via the graphical user interface, additional modification of the sketch from one or more additional users included in the social group.
4. The method of claim 1 , wherein the step of extracting further comprises:
generating a feature vector based on the one or more features from the image; and
mapping the feature vector in a feature-vector space, and wherein feature vectors corresponding to the one or more images are disposed within a threshold distance from the feature vector in the feature-vector space.
5. The method of claim 1 , wherein the encoding the modified sketch and processing the dimensional code are via a neural network, and wherein the neural network includes a generator and a discriminator, the generator comprising a first plurality of encoders and the plurality of decoders, each of which are arranged in a serial manner, and wherein the discriminator includes a concatenator, a second plurality of encoders, and a classifier.
6. The method of claim 5 , wherein the generator is configured to generate the image by sequentially encoding, via the first plurality of encoders, the modified sketch of the object, and sequentially decoding, via the plurality of decoders, an output of the first plurality of encoders.
7. The method of claim 5 , wherein the discriminator is configured for
concatenating, via the concatenator, the modified sketch of the object and the image to generate the concatenated image;
encoding, via the second plurality of encoders, the concatenated image; and
classifying, via the classifier, the encoded concatenated image.
8. A machine readable and non-transitory medium having information recorded thereon for searching content, wherein the information, when read by the machine, causes the machine to perform:
sharing, via a graphical user interface within a social group, a sketch of an object generated by a first user included in the social group;
receiving, via the graphical user interface, a modification to the sketch from at least a second user included in the social group;
encoding the modified sketch to generate a dimensional code reflecting dimensions of the sketch;
processing the dimensional code by a plurality of decoders to generate an image corresponding to the modified sketch of the object;
concatenating the modified sketch and the image to generate a concatenated image;
encoding the concatenated image;
determining a numeral value indicative of authenticity of the image by classifying the encoded concatenated image;
extracting, based on the authenticity, one or more features from the image;
identifying one or more images previously stored that have features similar to the one or more features; and
providing, on a webpage, the one or more images to the social group.
9. The medium of claim 7 , wherein the sketch of the object is drawn in the graphical user interface by the first user.
10. The medium of claim 9 , further receiving, via the graphical user interface, additional modification of the sketch from one or more additional users included in the social group.
11. The medium of claim 8 , wherein the step of extracting further comprises:
generating a feature vector based on the one or more features from the image; and
mapping the feature vector in a feature-vector space, and wherein feature vectors corresponding to the one or more images are disposed within a threshold distance from the feature vector in the feature-vector space.
12. The medium of claim 8 , wherein the encoding the modified sketch and processing the dimensional code are via a neural network, and wherein the neural network includes a generator and a discriminator, the generator comprising a first plurality of encoders and the plurality of decoders, each of which are arranged in a serial manner, and wherein the discriminator includes a concatenator, a second plurality of encoders, and a classifier.
13. The medium of claim 12 , wherein the generator is configured to generate the image by sequentially encoding, via the first plurality of encoders, the modified sketch of the object, and sequentially decoding, via the plurality of decoders, an output of the first plurality of encoders.
14. The medium of claim 12 , wherein the discriminator is configured for
concatenating, via the concatenator, the modified sketch of the object and the image to generate the concatenated image;
encoding, via the second plurality of encoders, the concatenated image; and
classifying, via the classifier, the encoded concatenated image.
15. A system for searching content, the system comprising:
a graphical user interface implemented by a processor and configured to
share, within a social group, a sketch of an object generated by a first user included in the social group, and
receive a modification to the sketch from at least a second user included in the social group;
a neural network implemented by the processor and configured to
encode the modified sketch to generate a dimensional code reflecting dimensions of the sketch,
process the dimensional code by a plurality of decoders to generate an image corresponding to the modified sketch of the object,
concatenate the modified sketch and the image to generate a concatenated image,
encode the concatenated image, and
determine a numeral value indicative of authenticity of the image by classifying the encoded concatenated image,
a feature extractor implemented by the processor and configured to extract, based on the authenticity, one or more features from the image; and
a processing unit implemented by the processor and configured to
identify one or more images previously stored that have features similar to the one or more features, and
provide, on a webpage, the one or more images to the social group.
16. The system of claim 15 , wherein the sketch of the object is drawn in the graphical user interface by the first user.
17. The system of claim 16 , wherein the sketch is further modified by one or more additional users included in the social group.
18. The system of claim 15 , wherein the feature extractor is further configured to
generate a feature vector based on the one or more features from the image; and
map the feature vector in a feature-vector space, and wherein feature vectors corresponding to the one or more images are disposed within a threshold distance from the feature vector in the feature-vector space.
19. The system of claim 15 , wherein the encoding the modified sketch and processing the dimensional code are via the neural network, and wherein the neural network includes a generator and a discriminator, the generator comprising a first plurality of encoders and the plurality of decoders, each of which are arranged in a serial manner, and wherein the discriminator includes a concatenator, a second plurality of encoders, and a classifier.
20. The system of claim 18 , wherein the generator is configured to generate the image by sequentially encoding, via the first plurality of encoders, the modified sketch of the object, and sequentially decoding, via the plurality of decoders, an output of the first plurality of encoders.