Trends, challenges and best practices for embedded vision

The challenges of lifecycle management for embedded vision systems

In recent years, there has been a growing trend in the use of embedded vision in a variety of industries such as automotive, healthcare, manufacturing and retail.

Embedded vision refers to the use of computer vision algorithms and technology on devices such as smartphones, drones or tiny system-on-chip (SoC) devices. Unlike a physical datacenter, such embedded devices have lower compute, limited storage and restricted power consumption. To address these challenges, some best practices have been developed, such as, for example, optimization of algorithms for specific hardware platforms.

Additionally, the development of machine learning and deep learning are helping to improve the accuracy and performance of embedded vision systems.

Gion-Pitchen Gross, Product Management & Marketing, Allied Vision, Oliver Helzle, Managing Director, Hema electronic, Orr Danon, CEO Hailo Technologies, and Olaf Munkelt, Co-founder, Co-owner and Managing Director, MVTec Software, discussed the latest trends in embedded vision with their host Ekaterina Sirazitdinova, Data Scientist, NVIDA, at the exhibitor forum of embedded world 2023.

“There is an ever-growing demand of identifying stuff”, Oliver Helzle answered the question why embedded vision systems are becoming more and more important. Embedded vision systems can handle fuzzy environments such as challenging outdoor conditions. Edge AI and thus embedded vision empowers you to look beyond the factory floor and enables you to respond quickly on changing circumstances Orr Danon and Gion-Pitchen Gross emphasised in their statements.

But of course, there are challenges to overcome in this area as well. Robust housing is one of the key words on the hardware side. “And on a long-term perspective you have to care for old equipment and maintain it to keep the systems running”, said Oliver Helzle. Especially in industries with very strict approval processes, such as the aviation industry, it is very important that embedded vision solutions have a service life of more than 2-3 years and can be maintained over a long period of time. Active lifecycle management is also very important on the software side. “You have to look at your old code again and again”, reportet software expert Olaf Munkelt.

Watch the whole panel discussion on embedded vision