- Case Study: Balancing the Demands of Algorithms and the Capabilities of Processors When Designing Computer Vision Systems
- Case Study: Good Decisions Require Good Information
- Case Study: How to Implement Deep Learning for Vision on Embedded Processors
- Case Study: How to Build Customer Confidence in a New Product
- Case Study: Growing Your Business by Targeting New Markets
Case Study: Making Sure Tools for Deep Learning are User-friendly and Robust
“…software is changing at head-spinning rates, driven by advances in how computers process giant amounts of data using artificial-intelligence techniques,” said the Wall Street Journal in a recent article, highlighting the challenge faced by smartphone software developers.
We agree, and it’s not just in smartphones. BDTI is seeing this trend across a wide range of applications. Visual intelligence, supported by sophisticated hardware and sensors, and delivered to users through increasingly complex software, is being deployed in automotive safety systems, agricultural equipment, unmanned aerial vehicles, retail analytics, and even trash collection. Because of the importance of vision-based applications, system designers increasingly consider the ease of software development as a key factor when choosing a processor for their designs. To be competitive, vendors must deliver robust and intuitive tools that enable developers to build efficient software easily.
A global semiconductor manufacturer, having invested heavily in deep learning, recently came to BDTI for an evaluation of its development kit for deep learning. Their objective was to ensure that the tools would enable typical application developers in specific markets to implement convolutional neural networks (CNNs) easily and efficiently, and that the resulting networks would provide the necessary accuracy. To test the use of the tools, BDTI defined a proxy application that models those found in the target market, then used the vendor’s tools to implement a CNN. Using its familiarity with a wide variety of tools for implementing neural networks, BDTI assessed the vendor’s tools, identifying strengths and weaknesses and prioritizing out the most important areas for improvement.
Discovery of a tool suite’s shortcomings prior to launch can save a product from disaster. And while one might think that the tool developers themselves should be able to see and fix problems, it’s difficult to perceive the way in which an application developer will use the tools. For almost 25 years BDTI's engineering team has delivered an independent perspective, resulting in tools that are user-friendly and robustly featured for application developers. In this most recent case, even though BDTI’s evaluation pointed out serious flaws, the vendor's tools manager was thrilled with the results – because it’s better to learn about problems before launch, when you can still do something about them
To learn more about BDTI's tool evaluation capabilities, contact Jeremy Giddings at BDTI (giddings@BDTI.com). Or, stop by BDTI’s exhibit in the Technology Showcase at the 2017 Embedded Vision Summit in Santa Clara, CA, May 1-3, where you can gain a deeper understanding of how computer vision is changing industries and business models, and learn about techniques and technologies for adding vision to many types of systems. Visit www.embeddedvisionsummit.com for details and registration. The Super Early Bird discount expires February 1, so secure your spot now using discount code nlid0130 and save 25%!