BenchmarkXPRT Blog banner

Tag Archives: object detection

Browser-based AI tests in WebXPRT 4: face detection and image classification

I recently revisited an XPRT blog entry that we posted from CES Las Vegas back in 2020. In that post, I reflected on the show’s expanded AI emphasis, and I wondered if we were reaching a tipping point where AI-enhanced and AI-driven tools and applications would become a significant presence in people’s daily lives. It felt like we were approaching that point back then with the prevalence of AI-powered features such as image enhancement and text recommendation, among many others. Now, seamless AI integration with common online tasks has become so widespread that many people unknowingly benefit from AI interactions several times a day.

As AI’s role in areas like everyday browser activity continues to grow—along with our expectations for what our consumer devices should be able to handle—reliable AI-oriented benchmarking is more vital than ever. We need objective performance data that can help us understand how well a new desktop, laptop, tablet, or phone will handle AI tasks.

WebXPRT 4 already includes timed AI tasks in two of its workloads: the “Organize Album using AI” workload and the “Encrypt Notes and OCR Scan” workload. These two workloads reflect the types of light browser-side inference tasks that are now fairly common in consumer-oriented web apps and extensions. In today’s post, we’ll provide some technical information about the Organize Album workload. In a future post, we’ll do the same for the Encrypt Notes workload.

The Organize Album workload includes two different timed tasks that reflect a common scenario of organizing online photo albums. The workload utilizes the AI inference and JavaScript capabilities of the WebAssembly (Wasm) version of OpenCV.js—an open-source computer vision and machine learning library. In WebXPRT 4, we used OpenCV.js version 4.5.2.

Here are the details for each task:

  • The first task measures the time it takes to complete a face detection job with a set of five 720 x 480 photos that we sourced from commercial photo sites. The workload loads a Caffe deep learning framework model (res10_300x300_ssd_iter_140000_fp16.caffemodel) using the commands found here
  • The second task measures the time it takes to complete an image classification job (labeling based on object detection) with a different set of five 718 x 480 photos that we sourced from the ImageNet computer vision dataset. The workload loads an ONNX-based SqueezeNet machine learning model (squeezenet.onnx v 1.0) using the commands found here.

To produce a score for each iteration of the workload, WebXPRT calculates the total time that it takes for a system to organize both albums. In a standard test, WebXPRT runs seven iterations of the entire six-workload performance suite before calculating an overall test score. You can find out more about the WebXPRT results calculation process here.

We hope this post will give you a better sense of how WebXPRT 4 measures one kind of AI performance. As a reminder, if you want to dig into the details at a more granular level, you can access the WebXPRT 4 source code for free. In previous blog posts, you can find information about how to access and use the code. You can also read more about WebXPRT’s overall structure and other workloads in the Exploring WebXPRT 4 white paper.

If you have any questions about this workload or any other aspect of WebXPRT 4, please let us know!

Justin

Potential web technology additions for WebXPRT 4

A few months ago, we invited readers to send in their thoughts and ideas about web technologies and workload scenarios that may be a good fit for the next WebXPRT. We’d like to share a few of those ideas today, and we invite you to continue to send your feedback. We’re approaching the time when we need to begin firming up plans for a WebXPRT 4 development cycle in 2021, but there’s still plenty of time for you to help shape the future of the benchmark.

One of the most promising ideas for WebXPRT 4 is the potential addition of one or more WebAssembly (WASM) workloads. WASM is a low-level, binary instruction format that works across all modern browsers. It offers web developers a great deal of flexibility and provides the speed and efficiency necessary for running complex client applications in the browser. WASM enables a variety of workload scenario options, including gaming, video editing, VR, virtual machines, image recognition, and interactive educational content.

In addition, the Chrome team is dropping Portable Native Client (PNaCL) support in favor of WASM, which is why we had to remove a PNaCL workload when updating CrXPRT 2015 to CrXPRT 2. We generally model CrXPRT workloads on existing WebXPRT workloads, so familiarizing ourselves with WASM could ultimately benefit more than one XPRT benchmark.

We are also considering adding a web-based machine learning workload with TensorFlow for JavaScript (TensorFlow.js). TensorFlow.js offers pre-trained models for a wide variety of tasks including image classification, object detection, sentence encoding, natural language processing, and more. We could also use this technology to enhance one of WebXPRT’s existing AI-themed workloads, such as Organize Album using AI or Encrypt Notes and OCR Scan.

Other ideas include using a WebGL-based workload to target GPUs and investigating ways to incorporate a battery life test. What do you think? Let us know!

Justin

Make confident choices about your company’s future tech with the XPRTs

Durham, NC, April 23, 2020 — Principled Technologies and the BenchmarkXPRT Development Community have released a video on the benefits of consulting the XPRTs before committing to new technology purchases.

AIXPRT, one of the battery of XPRT benchmark tools, runs image-classification and object-detection workloads to determine how well tech handles AI and machine learning.

CloudXPRT, another XPRT tool, accurately measures the end-to-end performance of modern, cloud-first applications deployed on infrastructure as a service (IaaS) platforms – allowing corporate decision-makers to select the best configuration for every objective.

All of the XPRTs give companies the real-world information necessary to determine which prospective future tech p – and which will disappoint

According to the video, “The XPRTs don’t just look at specs and features; they gauge a technology solution’s real-world performance and capabilities. So you know whether switching environments is worth the investment. How well solutions support machine learning and other AI capabilities. If next-gen releases beat their rivals or fall behind the curve.”

Watch the video at facts.pt/pyt88k5. To learn more about how AIXPRT, CloudXPRT, WebXPRT, MobileXPRT, TouchXPRT, CrXPRT, and HDXPRT can help IT decision-makers can make confident choices about future purchases, go to www.BenchmarkXPRT.com.

About Principled Technologies, Inc.
Principled Technologies, Inc. is the leading provider of technology marketing and learning & development services. It administers the BenchmarkXPRT Development Community.

Principled Technologies, Inc. is located in Durham, North Carolina, USA. For more information, please visit www.principledtechnologies.com.

Company Contact
Justin Greene
BenchmarkXPRT Development Community
Principled Technologies, Inc.
1007 Slater Road, Suite #300
Durham, NC 27703
BenchmarkXPRTsupport@PrincipledTechnologies.com

AIXPRT is here!

We’re happy to announce that AIXPRT is now available to the public! AIXPRT includes support for the Intel OpenVINO, TensorFlow, and NVIDIA TensorRT toolkits to run image-classification and object-detection workloads with the ResNet-50 and SSD-MobileNet v1networks, as well as a Wide and Deep recommender system workload with the Apache MXNet toolkit. The test reports FP32, FP16, and INT8 levels of precision.

To access AIXPRT, visit the AIXPRT download page. There, a download table displays the AIXPRT test packages. Locate the operating system and toolkit you wish to test and click the corresponding Download link. For detailed installation instructions and information on hardware and software requirements for each package, click the package’s Readme link. If you’re not sure which AIXPRT package to choose, the AIXPRT package selector tool will help to guide you through the selection process.

In addition, the Helpful Info box on AIXPRT.com contains links to a repository of AIXPRT resources, as well links to XPRT blog discussions about key AIXPRT test configuration settings such as batch size and precision.

We hope AIXPRT will prove to be a valuable tool for you, and we’re thankful for all the input we received during the preview period! If you have any questions about AIXPRT, please let us know.

AIXPRT Community Preview 3 is here!

We’re happy to announce that the AIXPRT Community Preview 3 (CP3) is now available! As we discussed in last week’s blog, testers can expect three significant changes in AIXPRT CP3:

  • We updated support for the Ubuntu test packages from Ubuntu version 16.04 LTS to version 18.04 LTS.
  • We added TensorRT test packages for Windows and Ubuntu. Previously, AIXPRT testers could test only the TensorFlow variant of TensorRT. Now, they can use TensorRT to test systems with NVIDIA GPUs.
  • We added the Wide and Deep recommender system workload with the MXNet toolkit for Ubuntu systems.


To access AIXPRT CP3, click this access link and submit the brief information form unless you’ve already done so for CP2. You will then gain access to the AIXPRT community preview page. (If you’re not already a BenchmarkXPRT Development Community member, we’ll contact you with more information about your membership.)

On the community preview page, a download table displays the currently available AIXPRT CP3 test packages. Locate the operating system and toolkit you wish to test, and click the corresponding Download link. For detailed installation instructions and information on hardware and software requirements for each package, click the corresponding Readme link. Instead of providing installation guide PDFs as we did for CP2, we are now directing testers to a public GitHub repository. The repository contains the installation readmes for all the test packages, as well as a selection of alternative test configuration files. We’ll discuss the alternative configuration files in more detail in a future blog post.

Note: Those who have access to the existing AIXPRT GitHub repository will be able to access CP3 in the same way as previous versions.

We’ll continue to keep everyone up to date with AIXPRT news here in the blog. If you have any questions or comments, please let us know.

Justin

Understanding the basics of AIXPRT precision settings

A few weeks ago, we discussed one of AIXPRT’s key configuration variables, batch size. Today, we’re discussing another key variable: the level of precision. In the context of machine learning (ML) inference, the level of precision refers to the computer number format (FP32, FP16, or INT8) representing the weights (parameters) a network model uses when performing the calculations necessary for inference tasks.

Higher levels of precision for inference tasks help decrease the number of false positives and false negatives, but they can increase the amount of time, memory bandwidth, and computational power necessary to achieve accurate results. Lower levels of precision typically (but not always) enable the model to process inputs more quickly while using less memory and processing power, but they can allow a degree of inaccuracy that is unacceptable for certain real-world applications.

For example, a high level of precision may be appropriate for computer vision applications in the medical field, where the benefits of hyper-accurate object detection and classification far outweigh the benefit of saving a few milliseconds. On the other hand, a low level of precision may work well for vision-based sensors in the security industry, where alert time is critical and monitors simply need to know if an animal or a human triggered a motion-activated camera.

FP32, FP16, and INT8

In AIXPRT, we can instruct the network models to use FP32, FP16, or INT8 levels of precision:

  • FP32 refers to single-precision (32-bit) floating point format, a number format that can represent an enormous range of values with a high degree of mathematical precision. Most CPUs and GPUs handle 32-bit floating point operations very efficiently, and many programs that use neural networks, including AIXPRT, use FP32 precision by default.
  • FP16 refers to half-precision (16-bit) floating point format, a number format that uses half the number of bits as FP32 to represent a model’s parameters. FP16 is a lower level of precision than FP32, but it still provides a great enough numerical range to successfully perform many inference tasks. FP16 often requires less time than FP32, and uses less memory.
  • INT8 refers to the 8-bit integer data type. INT8 data is better suited for certain types of calculations than floating point data, but it has a relatively small numeric range compared to FP16 or FP32. Depending on the model, INT8 precision can significantly improve latency and throughput, but there may be a loss of accuracy. INT8 precision does not always trade accuracy for speed, however. Researchers have shown that a process called quantization (i.e., approximating continuous values with discrete counterparts) can enable some networks, such as ResNet-50, to run INT8 precision without any significant loss of accuracy.

Configuring precision in AIXPRT

The screenshot below shows part of a sample config file, the same sample file we used for our batch size discussion. The value in the “precision” row indicates the precision setting. This test configuration would run tests using INT8. To change the precision, a tester simply replaces that value with “fp32” or “fp16” and saves the changes.

Config_snip

Note that while decreasing the precision from FP32 to FP16 or INT8 often results in larger throughput numbers and faster inference speeds overall, this is not always the case. Many other factors can affect ML performance, including (but not limited to) the complexity of the model, the presence of specific ML optimizations for the hardware under test, and any inherent limitations of the target CPU or GPU.

As with most AI-related topics, the details of model precision are extremely complex, and it’s a hot topic in cutting edge AI research. You don’t have to be an expert, however, to understand how changing the level of precision can affect AIXPRT test results. We hope that today’s discussion helped to make the basics of precision a little clearer. If you have any questions or comments, please feel free to contact us.

Justin

Check out the other XPRTs:

Forgot your password?