Image Recognition Systems

Driving Sophisticated Computer Vision with Machine Learning

One of the most powerful applications of machine learning and computer vision is image recognition. While it is trivial for a human being to recognize, categorize, or differentiate between images, the sheer volume of data encoded in a given image—especially one at a very high resolution—has until quite recently made computer-powered image recognition very challenging.

Today, image recognition systems powered by Neural Magic’s Inference Engine can be used to develop sophisticated computer vision applications for industries and use cases ranging from industrial manufacturing to retail to media—helping companies achieve previously impossible feats.

Bringing Computer Vision into its Own

Today, when machine learning engineers run image recognition models on a CPU, they often make sacrifices that affect the quality of outcomes, because they are forced to make untenable sacrifices between:

Performance
Batch Size
Accuracy
Cost

Neural Magic addresses these limitations by generating GPU-class performance on a CPU.

Software Flexibility

Bring computer vision to general-purpose CPUs. Run models anywhere: on premise, in the cloud, or at the edge.

Seamless Integration

Integrate seamlessly into existing CI/CD pipelines. Deploy in containers or virtual machines, and manage with Kubernetes like any modern software application.

Better Price Performance

High-performance algorithms that can run AI on commodity machines via smarter CPU memory utilization.

Currently supports image recognition models such as ResNet, MobileNet, VGGNet, EfficientNet

Learn More

Neural Magic In The News

Neural Magic Announces $15 Million in Seed Funding from Comcast Ventures, NEA, Andreessen Horowitz, Pillar VC and Amdocs

GPU Speeds without GPUs: Announcing the Neural Magic Inference Engine

Four Machine Learning Trends for Recommendation Systems

Try Neural Magic Today

The Neural Magic Inference Engine fits seamlessly into existing CI/CD pipelines, can be deployed in containers or virtual machines, and can be managed with Kubernetes like any modern software application.