2021 BRN Discussion, page-16998

I don't think Xailient have their own NN. This patent application talks about NNs, but it does nor describe them:

US2021150721A1 SYSTEMS AND METHODS FOR OBJECT DETECTION AND RECOGNITION

https://worldwide.espacenet.com/patent/search/family/075909198/publication/US2021150721A1?q=US2021150721A1

1 . A method for determining a zone of interest within an image, the method comprising:

receiving, by a computer processor, image data wherein the image data comprises a plurality of pixel groups, wherein each pixel group comprises an average distance therebetween such that the number of pixels within a smallest zone of interest falls between a minimum sample and a maximum sample within the potential zone of interest;

processing, by the computer processor, the plurality of pixel groups to determine a pixel value for pixels of at least one or more of the plurality of pixel groups, wherein the pixel value is determined from a comparison of plurality of pixel values for each pixel;

determining, by the computer processor, a pixel group value for each of the at least one or more of the plurality of pixel groups wherein the pixel group value is based on the pixel values of the pixels associated with the respective pixel group;

determining, by the computer processor, a heat map from the plurality of maximum pixel group values by way of a first neural network,

wherein the first neural network receives as input at least one pixel group value and produces as output the probability heat map comprising groups of graded values,

wherein the graded values are indicative of the probability that the respective pixel group includes a representation of an object of interest;

determining, by the computer processor, at least one zone of interest based on whether the groups of graded values meet a determined probability threshold; and

determining, by the computer processor, objects of interest within the at least one zone of interest by way of a second neural network,

wherein the second neural network receives as input the at least one zone of interest and produces as output a representation of the objects of interest,

wherein the representation of the objects of interest comprises at least a classification of each object of interest and the location thereof within the received image data.