Detection and identification of objects in images
Aspects of the disclosure provide for mechanisms for identification of objects in images using neural networks. A method of the disclosure includes: obtaining an image, representing each element of a plurality of elements of the image via an input vector of a plurality of input vectors, each input vector having one or more parameters pertaining to visual appearance of a respective element of the image, providing the plurality of input vectors to a first subnetwork of a neural network to obtain a plurality of output vectors, wherein each of the plurality of output vectors is associated with an element of the image, identifying, based on the plurality of output vectors, a sub-plurality of elements of the image as belonging to the image of the object, and determining, based on locations of the sub-plurality of elements, a location of an image of an object within the image.
1. A method, comprising:
obtaining an image, wherein the image comprises an image of an object (IO);
representing each superpixel of a plurality of superpixels of the image via an input vector of a plurality of input vectors, each input vector comprising one or more parameters pertaining to visual appearance of a respective superpixel of the plurality of superpixels of the image, wherein each superpixel comprises one or more pixels of the image;
providing, by a processing device, the plurality of input vectors to a first subnetwork of a neural network to obtain a plurality of output vectors, wherein each of the plurality of output vectors is associated with a superpixel of the plurality of superpixels of the image, and wherein an output vector associated with a first superpixel of the plurality of superpixels depends on input vectors of a predetermined subset of superpixels of the image adjacent to the first superpixel;
identifying, by the processing device, based on each output vector of a sub-plurality of the plurality of output vectors comprising a first component having a value that satisfies a threshold condition, a sub-plurality of the plurality of superpixels of the image as belonging to the IO; and
determining, by the processing device, based on locations of the sub-plurality of the plurality of superpixels, a location of the IO within the image.
2. The method of claim 1 , wherein each superpixel of the plurality of superpixels comprises one or more pixels obtained by rasterizing the image.
3. The method of claim 2 , wherein the one or more parameters pertaining to visual appearance of a superpixel of the image are determined in view of a first plurality of intensity values, wherein each of the first plurality of intensity values represents an intensity value for a first color for a respective pixel of the one or more pixels of the superpixel.
4. The method of claim 1 , wherein the one or more parameters pertaining to visual appearance of the superpixel of the image are determined by providing one or more input values pertaining to visual appearance of the superpixel of the image as input into a second subnetwork of the neural network.
5. The method of claim 4 , wherein each of the one or more input values pertaining to visual appearance of the superpixel of the image comprises an intensity value of at least one pixel of the superpixel.
6. The method of claim 4 , wherein the second subnetwork is a convolutional network.
7. The method of claim 3 , wherein the one or more parameters pertaining to visual appearance of the superpixel of the image are determined in further view of a second plurality of intensity values, wherein each of the second plurality of intensity values represents an intensity value for a second color for the respective pixel of the one or more pixels of the superpixel, wherein the second color is different from the first color.
8. The method of claim 1 , wherein the first subnetwork comprises a convolutional subnetwork or at least one hidden layer of neurons.
9. The method of claim 1 , wherein the first subnetwork was trained using a plurality of training images, wherein each image of the plurality of training images is prepared by:
obtaining a base image;
obtaining an image of a training object;
selecting a location for the image of the training object;
inserting the image of the training object at the selected location; and
adjusting a intensity value of a plurality of pixels of the inserted image in view of a reference intensity value of the base image.
10. The method of claim 9 , further comprising:
reducing quality of the inserted image of the training object.
11. A non-transitory machine-readable storage medium including instructions that, when accessed by a processing device, cause the processing device to:
obtain an image, wherein the image comprises an image of an object;
represent each superpixel of a plurality of superpixels of the image via an input vector of a plurality of input vectors, each input vector comprising one or more parameters pertaining to visual appearance of a respective superpixel of the plurality of superpixels of the image, wherein each superpixel comprises one or more pixels of the image;
provide the plurality of input vectors to a first subnetwork of a neural network to obtain a plurality of output vectors, wherein each of the plurality of output vectors is associated with a superpixel of the plurality of superpixels of the image, and wherein an output vector associated with a first superpixel of the plurality of superpixels depends on input vectors of a predetermined subset of superpixels of the image adjacent to the first superpixel;
identify, by the processing device, based on each output vector of a sub-plurality of the plurality of output vectors comprising a first component having a value that satisfies a threshold condition, a sub-plurality of the plurality of superpixels of the image as belonging to the TO; and
determine based on locations of the sub-plurality of the plurality of superpixels, a location of the IO within the image.
12. The non-transitory storage medium of claim 11 , wherein each superpixel of the plurality of superpixels comprises one or more pixels obtained by rasterizing the image.
13. A system comprising:
a memory; and
a processing device operatively coupled to the memory, the processing device to:
obtain an image, wherein the image comprises an image of an object;
represent each superpixel of a plurality of superpixels of the image via an input vector of a plurality of input vectors, each input vector comprising one or more parameters pertaining to visual appearance of a respective superpixel of the plurality of superpixels of the image;
provide the plurality of input vectors to a first subnetwork of a neural network to obtain a plurality of output vectors, wherein each of the plurality of output vectors is associated with a superpixel of the plurality of superpixels of the image, and wherein an output vector associated with a first superpixel of the plurality of superpixels depends on input vectors of a predetermined subset of superpixels of the image adjacent to the first superpixel;
identify, by the processing device, based on each output vector of a sub-plurality of the plurality of output vectors comprising a first component having a value that satisfies a threshold condition, a sub-plurality of the plurality of superpixels of the image as belonging to the IO; and
determine based on locations of the sub-plurality of the plurality of superpixels, a location of the IO within the image.