NorPix CUDA JPEG Compression Library Review: Specs, Price and Applications
Updated: 7 days ago
NorPix's CUDA JPEG library is a stand-alone C/C++ library that compresses camera images to JPEG in real time on NVIDIA GPUs, with load balancing across several GPUs. Pricing is available on request for the required configuration.
Quick Verdict
On an NVIDIA GTX 980 it reaches about 1.5 billion monochrome or 500 million colour pixels per second at quality 1 to 99%. In a StreamPix setup it allowed ten GigE cameras at 1280 x 1024 and 50 fps to record for up to 175 hours each on a 4 TB drive, 10 to 15 times longer than uncompressed.
NorPix CUDA JPEG Compression Library is most relevant when its stated function matches the intended process. This product belongs on the shortlist when its stated vision function matches the inspection or perception problem and the surrounding acquisition and control system can support it. Confirm the complete configuration and test representative data before scaling to production. A team seeking a turnkey station should evaluate the whole solution, including hardware, software licences, calibration and support.
CUDA JPEG at a Glance
Manufacturer: NorPix
Model: CUDA JPEG
Product type: GPU image compression library
Throughput (GTX 980): 1.5 Gpixel/s mono, 500 Mpixel/s colour
Quality: 1 to 99%
GPUs: Multiple NVIDIA GPUs, load balanced
Cameras: USB3 Vision, GigE, 10 GigE, Camera Link, CoaXPress
Delivery: Stand-alone library and run-time
CUDA JPEG Technical Specifications
Specification | NorPix CUDA JPEG Compression Library |
Manufacturer | NorPix |
Model | CUDA JPEG |
Product type | GPU image compression library |
Throughput (GTX 980) | 1.5 Gpixel/s mono, 500 Mpixel/s colour |
Quality | 1 to 99% |
GPUs | Multiple NVIDIA GPUs, load balanced |
Cameras | USB3 Vision, GigE, 10 GigE, Camera Link, CoaXPress |
Delivery | Stand-alone library and run-time |
Price | Price on request |
Design and Build Quality
A vision workflow should be assessed as a chain: image acquisition, processing and the decision delivered to the machine. The hardware or software can contribute useful capabilities without solving every part of that chain. Define which outputs the process needs and which conditions make a result unreliable. This keeps a successful demonstration from being confused with a validated inspection system.
Protection and Installation
Begin with representative images or measurements from the actual process. Include normal variation, difficult materials and borderline cases. Lighting, lens choice, camera position and data quality can affect the result as strongly as the processing method. Keep the acquisition conditions consistent while evaluating parameters, then deliberately test variation before committing to the installation.
Programming and Software
The appropriate workflow depends on whether the product supplies a graphical environment, a library, a camera interface or a complete measurement station. Use its documented interfaces for the selected configuration. A graphical setup can reduce initial programming, while a library can provide greater control inside an existing application. Neither removes the need for testing, configuration management and predictable recovery.
Integration and Operating Workflow
Separate configuration of the vision task from integration with the machine sequence. The application needs to associate a result with the correct workpiece and handle invalid data, delays and disconnects. Establish a clear acknowledgement between the processing system and its consumer so that an old or incomplete result cannot be mistaken for the next part's measurement.
CUDA JPEG Applications and Use Cases
Long-duration multi-camera recording
NorPix CUDA JPEG Compression Library is described for long-duration multi-camera recording. For this application, define the acceptance criterion before optimizing the setup. A useful evaluation includes both successful and unsuccessful cases and records how often the system makes the wrong decision. Measurement, detection and identification are different tasks; choose the output that the process actually needs rather than relying on an attractive visualization.
Bandwidth reduction
NorPix CUDA JPEG Compression Library is described for bandwidth reduction. For this application, define the acceptance criterion before optimizing the setup. A useful evaluation includes both successful and unsuccessful cases and records how often the system makes the wrong decision. Measurement, detection and identification are different tasks; choose the output that the process actually needs rather than relying on an attractive visualization.
Relevant Industries
Research
Industrial testing
Sports analysis
CUDA JPEG Price and Value
Price on request. Request a current quotation from NorPix for the intended configuration and deployment. No fixed USD price is assumed for this review.
The total vision-system budget can include cameras, optics, illumination, acquisition hardware, software licences, computing, calibration and integration. Ask for a quotation that identifies the included components and the intended deployment model. Development access, a runtime licence and a complete installed station can be different purchases.
Pros and Cons
Pros
Large storage savings
Scales with extra GPUs
Cons
Requires NVIDIA GPU
Who Should Choose CUDA JPEG?
This product belongs on the shortlist when its stated vision function matches the inspection or perception problem and the surrounding acquisition and control system can support it. Confirm the complete configuration and test representative data before scaling to production. A team seeking a turnkey station should evaluate the whole solution, including hardware, software licences, calibration and support.
CUDA JPEG FAQ
What is NorPix CUDA JPEG Compression Library?
NorPix's CUDA JPEG library is a stand-alone C/C++ library that compresses camera images to JPEG in real time on NVIDIA GPUs, with load balancing across several GPUs.
How much does CUDA JPEG cost?
Price on request. Obtain a current quotation for the required configuration, licences, accessories and services. Prices are indicative and may change.
What applications is CUDA JPEG intended for?
Its stated applications include Long-duration multi-camera recording, Bandwidth reduction. Validate the complete intended task before deployment.
What are the main specifications?
Throughput (GTX 980): 1.5 Gpixel/s mono, 500 Mpixel/s colour; Quality: 1 to 99%; GPUs: Multiple NVIDIA GPUs, load balanced; Cameras: USB3 Vision, GigE, 10 GigE, Camera Link, CoaXPress.
How should it be integrated?
Separate configuration of the vision task from integration with the machine sequence. The application needs to associate a result with the correct workpiece and handle invalid data, delays and disconnects. Establish a clear acknowledgement between the processing system and its consumer so that an old or incomplete result cannot be mistaken for the next part's measurement.
What are its main strengths?
Large storage savings Scales with extra GPUs
What should buyers check before selecting it?
Requires NVIDIA GPU Begin with representative images or measurements from the actual process. Include normal variation, difficult materials and borderline cases. Lighting, lens choice, camera position and data quality can affect the result as strongly as the processing method. Keep the acquisition conditions consistent while evaluating parameters, then deliberately test variation before committing to the installation.
Who is the best fit for CUDA JPEG?
This product belongs on the shortlist when its stated vision function matches the inspection or perception problem and the surrounding acquisition and control system can support it. Confirm the complete configuration and test representative data before scaling to production. A team seeking a turnkey station should evaluate the whole solution, including hardware, software licences, calibration and support.
Prices and package contents are indicative and may change. Confirm the final configuration, licensing, accessories, shipping, taxes and integration scope in a supplier quotation.
