Embedded deep learning perception software designed for the automotive market
Brodmann17 creates highly efficient deep-learning products, while maintaining state-of-the-art quality. The neural-network models are based on two main technological breakthroughs that have been invented and perfected by our research team over the past few years.
Brodmann17's novel neural-network inference
The world’s fastest neural-network
Brodmann17 has engineered core deep-learning algorithms from the ground-up and patented a new neural-network technology, that is based on weights and calculation sharing. Brodmann17 deep neural networks (NN) leverage this technology to minimize the number of unique calculations within the NN so that a significant number of recurring calculations may be shared between neurons. The result is a NN with identical accuracy, however with a much smaller number of calculations, enabling faster and lower power-compute networks compared to the standard NN approach.
standard solution
Standard single large pruned neural network
Brodmann17's patented solution
Weights & compute sharing neural-network
The Main Advantages
Significant advantage in runtime & power
Software only solution for any processor
State-of-the-art accuracy even on low power processors
Visit our Github repository to learn more
The world’s first automated training platform:
designed for automotive-grade
Brodmann17 has developed a unique platform in order to swiftly bring automated driving to the mass market.
Brodmann17 developed this comprehensive platform that ensures the best neural-network for any given challenge and processing power by optimizing the entire neural-network training process based on our IP: the data selection, parameter tuning, neural network architecture search (NAS) and deployment to target embedded processors for runtime benchmarks.
scaling Brodmann17’s operations & improving complex processes
The training platform was developed as a neural network production line to scale Brodmann17’s operations and improve the highly complex process of training and NN deployment that cannot be done at scale manually.
Performs methodological searches over thousands of servers simultaneously
The training platform performs methodological searches and continuously
re-iterates them over thousands of servers simultaneously. As it is cloud-based, the platform is infinitely scalable, and the training process can be completed as quickly as necessary.
Reducing time to market & costs
The training platform’s automation removes humans and lowers the associated risks of human error, while reducing time to market and costs.
cloud based, no privacy issues
The training platform is cloud based and access can be given to Brodmann17 partners to control and monitor their training process and solve privacy issue.
