Segmentation of data
Disclosed is a computer-implemented method for segmenting input data. In the method a plurality of tags is generated; the input data is masked with the plurality of tags; a plurality of output reconstructions is generated by inputting the plurality of masked input data to one of the following: a denoising neural network, a variational autoencoder; a plurality of values representing distances of each plurality of output reconstructions to the input data are determined; a plurality of updated versions of input data is generated by applying at least one of the determined values representing distances of each plurality of output reconstructions to the input data; and updated output reconstructions are generated by inputting the plurality of updated versions of input data to one of the networks. Also disclosed is a method for training the network and a processing unit.
1. A non-transitory computer-readable medium on which is stored a computer program which, when executed by a computer, performs a method for segmenting input data comprising:
generating tags for at least one data element in the input data;
masking the input data with the tags to generate masked input data, wherein the masked input data includes multiple portions of the input data that are segmented according to the tags;
generating output reconstructions by inputting the masked input data to a neural network;
determining values representing distances of each of the output reconstructions to the input data;
generating updated versions of input data by applying at least one of the values representing distances of each of the output reconstructions to the input data; and
generating updated output reconstructions by inputting the updated versions of input data to the neural network.
2. The non-transitory computer-readable medium of claim 1 , wherein generating the updated versions of input data comprises:
determining a minimum distance among the values representing the distances of each of the output reconstructions to the input data and a respective one of the tags that generated the minimum distance;
updating the generated tags by replacing the tags with the tag that generated the minimum distance; and
obtaining data elements of the input data corresponding to original tags from input data and other data elements from the output reconstruction data generated with the tag that generated the minimum distance for generating new input data.
3. The non-transitory computer-readable medium of claim 1 , wherein the method further comprises:
obtaining the values representing distances of each of the output reconstructions to the input data and providing them to the neural network for generating updated tags.
4. The non-transitory computer-readable medium of claim 1 , wherein the method further comprises:
obtaining the values representing distances of each of the output reconstructions to the input data and providing them to the neural network for generating updated tags;
determining a minimum distance among the values representing the distances of each of the output reconstructions to input data and the tag that generated the minimum distance; and
obtaining data elements of the input data corresponding to original tags from input data and other data elements from the output reconstruction data generated with the tag that generated the minimum distance for generating new input data.
5. The non-transitory computer-readable medium of claim 1 , wherein generating updated versions of input data comprises generating updated tags based on the values representing distances of each of the output reconstructions to the input data.
6. The non-transitory computer-readable medium of claim 1 , wherein the at least one data element includes at least one of a pixel or a group of pixels.
7. The non-transitory computer-readable medium of claim 1 , wherein the neural network is at least one of a denoising neural network or a variational autoencoder.
8. A method for segmenting input data, comprising:
generating tags for at least one data element in the input data, wherein the tags represent a segmentation of the input data;
masking the input data with the tags to generate masked input data, wherein the masked input data includes multiple portions of the input data that are segmented according to the tags;
generating output reconstructions by inputting the masked input data to a neural network;
determining values representing distances of each of the output reconstructions to the input data;
generating updated versions of input data by applying at least one of the values representing distances of each of the output reconstructions to the input data; and
generating updated output reconstructions by inputting the updated versions of input data to the neural network.
9. The method of claim 8 , wherein generating the updated versions of input data comprises:
determining a minimum distance among the values representing the distances of each of the output reconstructions to the input data and a respective one of the tags that generated the minimum distance;
updating the generated tags by replacing the tags with the tag that generated the minimum distance; and
obtaining data elements of the input data corresponding to original tags from input data and other data elements from the output reconstruction data generated with the tag that generated the minimum distance for generating new input data.
10. The method of claim 8 , wherein the method further comprises:
obtaining the values representing distances of each of the output reconstructions to the input data and providing them to the neural network for generating updated tags.
11. The method of claim 8 , wherein the method further comprises:
obtaining the values representing distances of each of the output reconstructions to the input data and providing them to the neural network for generating updated tags;
determining a minimum distance among the values representing the distances of each of the output reconstructions to input data and the tag that generated the minimum distance; and
obtaining data elements of the input data corresponding to original tags from input data and other data elements from the output reconstruction data generated with the tag that generated the minimum distance for generating new input data.
12. The method of claim 8 , wherein generating updated versions of input data comprises generating updated tags based on the values representing distances of each of the output reconstructions to the input data.
13. The method of claim 8 , wherein the at least one data element includes at least one of a pixel or a group of pixels.
14. The method of claim 8 , wherein the neural network is at least one of a denoising neural network or a variational autoencoder.
15. A system for segmenting input data, comprising:
a memory; and
a processor that is operable to execute instructions that are stored in the memory, wherein the instructions, when executed by the processor, cause the processor to:
generate tags for at least one data element in the input data, wherein the tags represent a segmentation of the input data;
mask the input data with the tags to generate masked input data, wherein the masked input data includes multiple portions of the input data that are segmented according to the tags;
generate output reconstructions by inputting the masked input data to a neural network;
determine values representing distances of each of the output reconstructions to the input data;
generate updated versions of input data by applying at least one of the values representing distances of each of the output reconstructions to the input data; and
generate updated output reconstructions by inputting the updated versions of input data to the neural network.
16. The system of claim 15 , wherein the instructions to generate the updated versions of input data further cause the processor to:
determine a minimum distance among the values representing the distances of each of the output reconstructions to the input data and a respective one of the tags that generated the minimum distance;
update the generated tags by replacing the tags with the tag that generated the minimum distance; and
obtain data elements of the input data corresponding to original tags from input data and other data elements from the output reconstruction data generated with the tag that generated the minimum distance for generating new input data.
17. The system of claim 15 , wherein the instructions further cause the processor to:
obtain the values representing distances of each of the output reconstructions to the input data and provide them to the neural network for generating updated tags.
18. The system of claim 15 , wherein the instructions further cause the processor to:
obtain the values representing distances of each of the output reconstructions to the input data and provide them to the neural network for generating updated tags;
determine a minimum distance among the values representing the distances of each of the output reconstructions to input data and the tag that generated the minimum distance; and
obtain data elements of the input data corresponding to original tags from input data and other data elements from the output reconstruction data generated with the tag that generated the minimum distance for generating new input data.
19. The system of claim 15 , wherein generating updated versions of input data comprises generating updated tags based on the values representing distances of each of the output reconstructions to the input data.
20. The system of claim 15 , wherein the at least one data element includes at least one of a pixel or a group of pixels.
21. The system of claim 15 , wherein the neural network is at least one of a denoising neural network or a variational autoencoder.