We Accelerate Convolutional Neural Networks
Don Gaspar and Mark Mathews founded Gigantor Technologies to make Machine Vision neural network performance real-time, reduce latency, and offer HD, 4K, 8K and beyond frame sizes for your inference models.
The results of our efforts is GigaMACS™, an AI Accelerator that will take an existing TensorFlow or other Neural Network model and creates a high-performance optimized model out of it for use with an FPGA or ASIC.
GigaMACS™ accelerates Neural Network models to perform as fast as your input and up to real-time, and is perfrect for machine vision, radar, lidar and other devices requiring low-latency, large frame sizes, and high speeds.
Have a faster camera, radar, lidar? GigaMACS™ will ensure your model can keep up with the input rate without losing any data. Regardless of model complexity or workload, GigaMACS™ will make it perform real-time and near-zero latency.
Only a Tardis Manages your Clock Better
Adding faster clocks and more memory to speed up your Neural Network is not going to help complicated models that require real-time, low-latency, large frame sizes and high-speed performance. The problem with GPUs and TPUs is architectural, and general computing is great but is not designed for inference when it was envisioned 70 years ago, so these approaches will never achieve high-speed deterministic throughput with low-latency. A new non-von Neumann approach was taken with GigaMACS™ to solve all these issues. The results speak for themselves.
GigaMACS™ can optimize your existing Neural Network model to new production heights while helping you migrate to an ASIC for the ultimate performance. Only a Timelord from Gallifrey can make a Neural Network leverage time better.