How Tecniplast is Using Edge AI to Digitize the Reluctant Market

Introducing digital transformation to industries inherently resistant poses challenges, so we welcomed discussion with Tecniplast. The ST customer designs cages and enclosures for life science research and healthcare. Based in Italy, Tecniplast recently unveiled a Digital Ventilated Cage leveraging capacitive sensors, computer vision and an STM32H7-powered neural network.

By digitizing data collection instead of physical inspections, researchers avoid disturbing rodents under study and consequent impacts on behavioral results. Access to more detailed, remote monitoring has opened new areas of discovery.

Where oversight had disincentivized tech adoption due to risk of compromising delicate experiments, Tecniplast demonstrated digital solutions need not introduces such tradeoffs when carefully engineered. Their innovation upheld research integrity while reaping analytics benefits – a nuanced example of positively navigating digital shifts even in risk-averse domains.

Such customer discussions provide valuable perspective on applying embedded technologies where circumstances require especially sensitive innovation. Tecniplast shows with targeted design, digital tools need not undermine vital considerations and may expand what’s possible.

How to face a reluctant market?

Tecniplast's use case highlights clear upsides to digitization. However, as they described, most cage manufacturers and end users currently favor analog methods. Their digital ventilated cage thus stands out, as the industry at large appears resistant to adopting sensors and microcontrollers.

Some lack know-how around initiating such developments, while others remain comfortable with existing approaches. Additionally, the enormous installed base of conventional cages is leveraged as a rationale against transitioning platforms.

Therefore, to justify the change demanded by a digital solution, it must offer unequivocal benefits while maintaining cost competitiveness. Significant advantages and cost-effectiveness are necessities to overcome entrenched preferences and millions of functional analog units currently deployed.

Tecniplast's challenge exemplifies how digitization proponents in traditional fields must address not only technical merits but also overcome behavioral inertia. Considerable performance gains and value at comparable expense are understandably key persuasive factors in effectively catalyzing paradigm shifts.

How to solve challenges by shipping a first solution?

To solve these multidimensional challenges, Tecniplast started with low-risk experimentation. They installed capacitive plates under cages and utilized off-the-shelf camera systems, keeping all sensors external to ensure simplicity, cost control, non-disruption of animal behavior and accident prevention. Plates tracked rodent movement day and night while cameras monitored consumables. Due to distinguishing feed levels from obstructions better than time-of-flight alternatives, image CMOS proved most reliable.

Valuable lessons came from their first DVC's use. Excess data collection skyrocketed cloud processing expenses. While computer vision showed promise, image sensors proved unreliable. Sending JPEG data to cloud also lagged and inflated costs.

This drove investigating local neural network processing. Coinciding with an ST keynote addressing on-device analytics, Tecniplast recognized potential using STM32s. Leveraging embedded ML minimized data transfers while improving reliability, performance and economics versus previous cloud-centric solutions.

Through iterative refinement guided by operational feedback, Tecniplast enhanced their solution's attributes most critical to stakeholders – scientific integrity, economic feasibility and user convenience.

Spoiler: What were they able to accomplish?

By addressing issues identified through initial testing, Tecniplast unlocked unprecedented insights. Researchers now attain exact water monitoring capabilities never seen before. Technicians can detect rodent drinking sessions and intake volumes, raising flags much sooner if anomalies surface.

Previously, gathering such intel proved infeasible. As nocturnal drinkers starting after 6pm, cage observation through the night would have mandated constant human presence necessarily impacting rodent behavior. Now automated recording eliminates observation biases while capturing new nocturnal consumption patterns.

Remote tracking at this granularity has opened doors to studying previously unknown behavioral attributes only visible through continuous digital documentation devoid of human interaction effects.

Solving reliability and data transfer challenges afforded discovery opportunities beyond even Tecniplast's goals. Their innovations now empower scientific examinations into new natural domains previously left unexplored due to technical barriers inevitably tying research to direct oversight that risks impacting authentic behavior.

How is it going?

How did Tecniplast gravitate toward ST?

While local processing was imperative, Tecniplast also knew external memory inflated costs beyond viability. As convincing resistance demands revolutionary benefits at parity expenses, local solutions were paramount.

They evaluated STM32MP1 and STM32H7 suitability. Could image analysis leverage a microprocessor, or had microcontrollers grown sufficiently potent? Experimenting with our B-CAMS-OMV, Tecniplast tested JPEG handling of MPU and MCU hardware IPs.

Rather than raw video's exorbitant bandwidth and processing demands, the company extracted and compressed frames. Interestingly, STM32H7 emerged as the preferred choice meeting computational needs within cost constraints without extra modules.

By leveraging our solutions, Tecniplast gained insight validating microcontroller sufficiency for their application's demands. Locally executing full operations on a single chip answered performance requirements while satisfying economic necessity in this sensitive industry.

Such customer collaboration helps confirm technology viability through real-world testing that can unlock new embedded applications in resistant markets requiring revolutionary yet responsible innovation.

How did ST help Tecniplast with machine learning?

The domain stm32cube.ai is currently for sale for €1000
The owner of the domain has listed it for sale on Sedo Domain Parking
Sedo maintains no direct relationship with third party advertisers and does not endorse any services/trademarks listed on the domain parking page
The domain privacy policy can be found by clicking the “Privacy Policy” link
Some potential next steps if interested in purchasing the domain:

Review Sedo's purchase process and policies on domain parking pages
Consider if the asking price of €1000 is reasonable based on the domain value
Check if the domain owner/Sedo allow making an offer below the listed price
Review Sedo's ownership verification and transfer procedures
Evaluate what plans you have for the domain if purchased
The domain appears to be a relevant one for STM32 and embedded development resources given the keywords included. Its sale listing could be an opportunity for the right buyer depending on their intended usage and budget.

What’s next?

Tecniplast's DVC is gaining meaningful market traction, driving transformative mindset shifts. Increased cage shipments coincide with teams already planning their next innovations.

Now that these smart cages have proven so effectively powerful, Tecniplast explores monitoring additional behaviors like postures, feeding routines, movements and beyond. Leveraging more sensors and powerful MCUs/MPUs via evolving neural algorithms, the aim expands researchers’ observational breadth.

Simply put, Tecniplast occupies a distinctive industry position and seeks leveraging this advantage through continuous betterment. Rather than superficial digital replication of analog processes, the goal illuminates new frontiers benefitting academics and biomedicine.

By demonstrating digitalization’s potential for groundbreaking insight over incremental process digitization, Tecniplast provides a model inspiring broader acceptance of technology’s role in scientific progress. Their visionary work cultivates understanding that embedded intelligence appropriately applied cultivates discovery.

Growing adoption marks critical recognition of Tecniplast’s innovation while their aspirations pursue an even brighter future for life science research through ever more sophisticated yet respectful digital tools.


About The Author

Ibrar Ayyub

I am an experienced technical writer with a Master's degree in computer science from BZU Multan University. I have written for various industries, mainly home automation, and engineering. I have a clear and simple writing style and am skilled in using infographics and diagrams. I am a great researcher and is able to present information in a well-organized and logical manner.

Follow Us:
LinkedinTwitter
Scroll to Top