Case Study: Balancing the Demands of Algorithms and the Capabilities of Processors When Designing Computer Vision Systems

Submitted by BDTI on Thu, 12/15/2016 - 00:00

There is now little question that deep learning is an effective means for a wide range of detection and recognition applications. It is increasingly used in computer vision, where it has vastly improved accuracy rates for object recognition. In some cases, the effectiveness of deep learning has also resulted in unrealistic—or, as-yet unrealizable—expectations. For example, one customer came to BDTI with a long list of detection and recognition functions that it wanted implemented on a low-cost embedded processor using a neural network.

BDTI’s challenge was to balance the computational demands of neural networks, the broad set of customer requirements, and the constraints of the customer’s processor. As its first step, BDTI asked a series of questions to clarify the customer’s requirements and then performed an analysis of algorithm options. BDTI then identified an algorithm that would both meet the customer’s functionality requirements and fit on the target processor—a low-cost embedded processor without any specialized neural network engine. Finally, BDTI carefully optimized the algorithm to use both the CPU and GPU efficiently, enabling real-time operation.  The key to the success of this effort was BDTI’s strong knowledge of processor architectures and deep understanding of artificial neural networks.

The effectiveness of neural networks for a growing number of applications, from automotive and industrial safety to medical monitoring, is undeniable. One key to a successful product is an understanding of the interplay between algorithms and processor architectures. If you are considering the use of neural networks in your application, BDTI can help with processor selection, algorithm design and implementation. Contact Jeremy Giddings ( for a confidential discussion of your requirements.

Come see demos of BDTI implementations of computer vision and deep learning at the 2017 Embedded Vision Summit in Santa Clara, CA, May 1-3. And, to understand how computer vision is changing industries and business models, and to learn about techniques and technologies for adding vision to many types of systems, attend the Summit conference program. Visit for details and registration.

Add new comment

Log in or register to post comments