TomTom has filed a patent for a computer processing system that trains a model for semantic image segmentation. The model includes a refinement neural network and a discriminator neural network. The refinement neural network generates predicted segmentation maps using random values and outputs them to the discriminator neural network. The system trains the refinement neural network using an objective function that includes a term representing the difference between the predicted label distribution and the average of the predicted segmentation maps. GlobalData’s report on TomTom gives a 360-degree view of the company including its patenting strategy. Buy the report here.
Smarter leaders trust GlobalData
According to GlobalData’s company profile on TomTom, AV localization algorithms was a key innovation area identified from patents. TomTom's grant share as of June 2023 was 1%. Grant share is based on the ratio of number of grants to total number of patents.
The patent is filed for a computer system for semantic image segmentation
A recently filed patent (Publication Number: US20230186100A1) describes a computer system that is capable of training a model for semantic image segmentation. The model consists of a discriminator neural network and a refinement neural network. The refinement neural network receives a predicted label distribution for an image and generates multiple predicted segmentation maps using random values obtained from a noise source. These predicted segmentation maps are then passed to the discriminator neural network. The computer system is configured to train the refinement neural network using an objective function that includes a term representing the difference between the predicted label distribution and the average of the predicted segmentation maps.
The computer system described in the patent is designed to implement the model for semantic image segmentation. It includes one or more processors that are responsible for training the refinement neural network and the discriminator neural network. The refinement neural network receives the predicted label distribution from a calibration neural network, which is also part of the model. The calibration neural network takes input images and produces predicted label distributions using a likelihood-based semantic segmentation approach. The computer system is capable of training both the calibration neural network and the discriminator neural network.
The training process for the refinement neural network and the discriminator neural network involves a generative-adversarial-network (GAN) training process. The computer system conditions both neural networks on the input image during training. The discriminator neural network is trained using a loss function that measures its ability to discriminate between the predicted segmentation maps generated by the refinement network and ground-truth label data.
The objective function used to train the refinement neural network is a combination of two terms. The first term depends on the output of the discriminator network, while the second term depends on the difference between the predicted label distribution and the average of the predicted segmentation maps generated by the refinement network. The average of the predicted segmentation maps is calculated by taking the pixel-wise arithmetic mean of the multiple predicted segmentation maps.
The computer system employs a gradient descent process to train any of the neural networks involved in the model. The patent also describes a method for training the model, as well as a non-transitory computer-readable storage medium that stores instructions for implementing the training and implementation of the model for semantic image segmentation.
To know more about GlobalData’s detailed insights on TomTom, buy the report here.
Data Insights
From
The gold standard of business intelligence.
Blending expert knowledge with cutting-edge technology, GlobalData’s unrivalled proprietary data will enable you to decode what’s happening in your market. You can make better informed decisions and gain a future-proof advantage over your competitors.
GlobalData, the leading provider of industry intelligence, provided the underlying data, research, and analysis used to produce this article.
GlobalData Patent Analytics tracks bibliographic data, legal events data, point in time patent ownerships, and backward and forward citations from global patenting offices. Textual analysis and official patent classifications are used to group patents into key thematic areas and link them to specific companies across the world’s largest industries.
