Nota developed and owned the core technology for AI model compression. We at Nota are continuously focusing on core technologies and Intellectual Properties related to AI model compression. Nota builds AI models that compact, accurate, and optimized per each device. Achieving unparalleled excellence comes from the accuracy that are Nota values the most. Nota's technological differentiation is its capability to deliver accuracy with the lighter and faster models.
Intellectual property rights 25+
Published papers 10+
NetsPresso is the automatic AI compression platform that quickly maximizes the AI model compression rate while minimizing the loss of performance to run on edge devices with or without network.
Input
Accommodates various types of pre-trained models and enables setting requirements for latency, size, and etc. to meet the customer needs.
- Pruning
- Quantization
- Knowledge distillation
- Neural architecture search
- Filter decomposition
- etc.
Compression
methods
Supports a variety of methods for AI model compression. Per each method, can set parameters to optimize AI model compression flexibly per data and target devices in order to accommodate diverse service requirements.
- Pruning
- Quantization
- Knowledge distillation
- Neural architecture search
- Filter decomposition
- etc.
- Rpi
- Coral board
- NVIDIA jetson
- Mobile
- MCU
HPO
What makes NetsPresso special is the optimized compression parameters by testing directly on the target devices where the AI model will be deployed. With a pool of various devices including Jetson Nano and Raspberry Pi for testing, AI models become truly optimized.
- Rpi
- Coral board
- NVIDIA jetson
- Mobile
- MCU
Output
Through the AI Model compression pipeline of NetsPresso, we complete creating a stand-alone AI model that is compact, light and accurate.
Lighter, faster, and accurate AI models come from NetsPresso.
NetsPresso is unique as it does not compromise accuracy even with all these advantages. The compressed AI model improved its inference time up to 8 times faster and significantly reduced power consumption compared to the original model. NetsPresso can save 80% or more on product/service development and maintenance costs.
Nota does not compromise performance while compressing AI models.
Light and fast. Thanks to the effectively improved inference time, you can use AI without interruption.
Build the most cost-effective AI model. You can use AI models at affordable prices.
Can be used on edge devices for a long time through minimized power consumption.
Light but accurate AI, this is Nota’s technological differentiation. Existing model compression techniques must trade off between compression ratio and accuracy. But NetsPresso is different. Maximizing compression while maintaining accuracy is the core of NetsPresso.
Nota's compressed image classification model identifies a wide variety of objects including people, animals, plants and equipment.
Nota's compressed face authentication model can identify and authenticate up to 30,000 people on edge devices. Nota's face authentication system, which results in fast and accurate performance even with diverse lighting settings and personal protective equipment, can be used in various contactless situations including access control, membership authentication, driver recognition, etc.
Nota's compressed object detection model detects and locates various objects such as vehicles, people, and equipment on edge devices. Nota's object detection model, which ensures fast and accurate performance regardless of network availability, can be used in intelligent monitoring systems for security surveillance cameras and advanced driver assistance systems to provide contextual information.
Nota's compressed hand gesture recognition model can discern more than 10 different hand gestures on a miniaturized edge device equipped with a camera. Experience the contactless interface by using Nota's technology that can control smart devices in everyone’s daily lives.