System and method for multi-modal image classification
Systems and methods for classifying images (e.g., ads) are described. An image is accessed. Optical character recognition is performed on at least a first portion of the image. Image recognition is performed via a convolutional neural network on at least a second portion of the image. At least one class for the image is automatically identified, via a fully connected neural network, based on one or more predictions, each of the one or more predictions being based on both the optical character recognition and the image recognition. Finally, the at least one class identified for the image is output.
1. A computer-implemented method, comprising:
preprocessing an image to identify a plurality of elements from the image;
performing optical character recognition on at least a first element of the plurality of elements from the image to identify text in the first element;
converting the text in the first element into a term feature vector;
performing image recognition, via a convolutional neural network, on at least a second element of the plurality of elements from the image to identify an image feature;
predicting, via a fully connected neural network and a concatenation of the term feature vector and the image feature, (i) that an advertisement is present in the image, (ii) a category of the advertisement, and (iii) a brand based on the category of the advertisement; and
outputting the category of the advertisement.
2. The method of claim 1 , wherein preprocessing the image comprises:
eroding or dilating thickness of the plurality of elements; or
adjusting contrast in and around the plurality of elements.
3. The method of claim 1 , further comprising:
cleaning, via a natural language processing, the text in the first element.
4. The method of claim 1 , further comprising:
resizing the image such that the text in the plurality of elements is recognizable by the convolutional neural network.
5. The method of claim 4 ,
wherein the convolutional neural network determines edges or boundaries of the plurality of elements within the resized image.
6. The method of claim 5 , wherein the image feature comprises a logo of the brand.
7. The method of claim 1 , wherein:
the fully connected neural network comprises a plurality of layers, each of the layers comprising a plurality of artificial neurons,
each of the neurons performs one or more calculations using one or more parameters, and
each of the neurons of a first layer of the plurality of layers is connected to each of the neurons of a second layer of the plurality of layers.
8. The method of claim 7 , wherein:
the convolutional neural network comprises an input layer, a plurality of hidden layers, and an output layer,
the plurality of hidden layers comprises a convolutional layer, a rectified linear unit layer, and a pooling layer,
neurons of each of the layers transforms a three-dimensional input volume to a three-dimensional output volume of neuron activations, and
the neurons of one of the layers do not connect to all of the neurons of another of the layers.
9. The method of claim 1 , wherein preprocessing the image comprises manipulating the image via a function obtained from an image library.
10. The method of claim 1 , further comprising retrieving the image from a remote server.
11. The method of claim 1 , wherein the category of the advertisement is at least one of travel, automotive, finance, consumer goods, retail, restaurants, sports, entertainment, telecommunications, healthcare, insurance, computer technology, education, or business services.
12. A system comprising one or more processors coupled to one or more storage devices storing instructions that, when executed, cause the one or more processors to perform the following operations:
preprocessing an image to identify a plurality of elements from the image;
performing optical character recognition on at least a first element of the plurality of elements from the image to identify text in the first element;
converting the text in the first element into a term feature vector;
performing image recognition via a convolutional neural network, on at least a second element of the plurality of elements from the image to identify an image feature;
predicting, via a fully connected neural network and a concatenation of the term feature vector and the image feature, (i) that an advertisement is present in the image, (ii) a category of the advertisement, and (iii) a brand based on the category of the advertisement; and
outputting the category of the advertisement.
13. The system of claim 12 , wherein:
the fully connected neural network comprises a plurality of layers, each of the layers comprising a plurality of artificial neurons,
each of the neurons performs one or more calculations using one or more parameters, and
each of the neurons of a first layer of the plurality of layers is connected to each of the neurons of a second layer of the plurality of layers.
14. The system of claim 13 , wherein:
the convolutional neural network comprises an input layer, a plurality of hidden layers, and an output layer,
the plurality of hidden layers comprises a convolutional layer, a rectified linear unit layer, and a pooling layer,
neurons of each of the layers transforms a three-dimensional input volume to a three-dimensional output volume of neuron activations, and
the neurons of one of the layers do not connect to all of the neurons of another of the layers.
15. The system of claim 12 , wherein the category of the advertisement is at least one of travel, automotive, finance, consumer goods, retail, restaurants, sports, entertainment, telecommunications, healthcare, insurance, computer technology, education, or business services.
16. A non-transitory computer-readable medium comprising instructions executable by at least one processor to perform a method, the method comprising:
preprocessing an image to identify a plurality of elements from the image;
performing optical character recognition on at least a first element of the plurality of elements from the image to identify text in the first element;
converting the text in the first element into a term feature vector;
performing image recognition, via a convolutional neural network, on at least a second element of the plurality of elements from the image to identify an image feature;
predicting, via a fully connected neural network and a concatenation of the term feature vector and the image feature, (i) that an advertisement is present in the image, (ii) a category of the advertisement, and (iii) a brand based on the category of the advertisement; and
outputting the category of the advertisement.
17. The medium of claim 16 , wherein:
the fully connected neural network comprises a plurality of layers, each of the layers comprising a plurality of artificial neurons,
each of the neurons performs one or more calculations using one or more parameters, and
each of the neurons of a first layer of the plurality of layers is connected to each of the neurons of a second layer of the plurality of layers.
18. The medium of claim 17 , wherein:
the convolutional neural network comprises an input layer, a plurality of hidden layers, and an output layer,
the plurality of hidden layers comprises a convolutional layer, a rectified linear unit layer, and a pooling layer,
neurons of each of the layers transforms a three-dimensional input volume to a three-dimensional output volume of neuron activations, and
the neurons of one of the layers do not connect to all of the neurons of another of the layers.
19. The medium of claim 16 , wherein the category of the advertisement is at least one of travel, automotive, finance, consumer goods, retail, restaurants, sports, entertainment, telecommunications, healthcare, insurance, computer technology, education, or business services.