Giants of Neural Network Acceleration
Gigantor’s multi-patented product, GigaMACS™, is the solution for accelerating and moving latency to near-zero for CNN/DNN models in artificial intelligence.
GigaMACS™ accelerates convolutional and deep neural networks to process input pixels as fast as the source delivers images. The model will run with near-zero latency while accepting high-definition 4K input pixels at 240 frames per second or more. Your neural network model is not pruned, filtered, or have any mathematical precision reduced.
What would your business look like if your neural network model could accept HD/4K full-frame images at the camera speed and process outputs in real-time? For autonomous vehicles, this could save lives. For agricultural vehicles, it could mean doubling your speed and productivity. For military and aerospace, it could mean accurate targeting of tens of thousands of objects in real-time. For medical devices, it brings real-time diagnosis.
Click here to find out how our patented technology can accelerate your neural network to real-time without data loss, and reduce latency.
 |
 |
 |
||
| No Changes | No RAM | Near-zero Latency | ||
| No alterations or changes to the model in any way | No RAM is utilized, which eliminates access bottlenecks | Even while processing 240 FPS, latency is less than ½ msec | ||
 |
 |
 |
||
| Frame Resolution | High Speed | Saves Power | ||
| Process full size, HD, 4K, & 8K images without dropping frames |
Accepts input pixels as fast as the camera can deliver them, 60, 240, or 1,000 FPS and even quicker | Saves more than 90% of standard power consumption |
GigaMACS™ accelerates AI technology to operate in real-time with constant near-zero latency, for real-world applications, without dropping frames, while reducing power consumption. The novel solution accelerating convolutional and deep neural network models is unlike anything seen before. The US Patent Office allowed all our claims without contest on both patents: in only four (4) months, Patent 11,099,854 issued on 8/24/2021, and; in just seven (7) months, Patent 11,256,981 issued on 2/22/2022, and Patent 11,354,571 was issued on 06/07/2022.
|
We Accelerate Neural Networks For:
|
|