Enabling Intelligent Autonomy through highly-efficient AI vision processing
Recogni's purpose-built AI vision processing solutions make the benefits of real-time object detection and classification accessible across a wide range of industries. Our leading performance and low energy consumption enable you to address current problems and open new possibilities for innovation.
Intelligent Autonomy requires a deeper understanding of the environment to address critical challenges and open new opportunities.
Better vision processing is needed to automate manual tasks and enable innovation which can address the challenges of labor availability and productivity, operational downtime, resource conservation, environmental sustainability, and safe operation.
AI processing solutions require multiple image sensor inputs with high performance, low latency, and low energy use to avoid costly liquid cooling and extend operating range.
Farming
Mining
Trucking
Smart manufacturing
Delivery
Aircraft
Recogni's AI processing solutions are purpose-built to provide high performance at low power.
- Purpose-built for highest compute density and efficiency
- Multi-stream image sensor input with high performance and low latency
- Standard workflows and bit-accurate simulation
- Recogni model zoo with vision processing models
- No need for liquid cooling
- Lowest total system cost
10x
better power efficiency
<10ms
processing latency
5x
higher compute density
Recogni enables customers to open the opportunities of Intelligent Autonomy
Scorpio
Scalable from 150 to 1000 TFLOPS of AI compute
Processes multiple high-resolution image sensors at up to 30 frames per second
Under 10 ms latency from last-pixel-in to perception-out
Efficiency up to 40 TFLOPS per watt
Pegasus
AI inference card gives developers access to Scorpio's benefits
Compatible with x86 and ARM host processors
Standard form factor with four industry standard image sensor inputs
Enables rapid deployment of AI workloads
AI development platform
Provides flexibility, ease-of-integration, and the ability to scale
Bit-exact simulation
PTQ and QAT training methods to optimize models for best accuracy
Leverage Recogni model zoo and design guidelines
