cd /news/computer-vision/autoiq-ai-ai-powered-isp-tuning-on-s… · home › topics › computer-vision › article
[ARTICLE · art-171003] src=blog.st.com ↗ pub= topic=computer-vision verified=true sentiment=↑ positive

AUTOIQ.ai, AI-powered ISP tuning on STM32 and with STM32 middleware; the new benchmark in computer vision

Emmetra, an ST Partner Program member, has released AUTOIQ.ai, a sensor calibration and image quality optimization tool that uses AI to automatically tune image signal processor parameters on STMicroelectronics' STM32 hardware ISP, demonstrated on the STM32N6 microcontroller and STM32MP257 microprocessor via Windows and Linux desktop software. Emmetra said customers reported the tool reduced friction and helped them finalize designs faster, and a macOS version is expected next year. The tool targets teams that lack deep image-pipeline expertise or need application-specific indicators, such as AI vision, that general-purpose tools like STM32 ISP IQTune do not provide.

by read5 min views3 publishedNov 17, 2026
AUTOIQ.ai, AI-powered ISP tuning on STM32 and with STM32 middleware; the new benchmark in computer vision
Image: Blog (auto-discovered)

AUTOIQ.ai is a sensor calibration and image quality optimization tool from Emmetra, a member of the ST Partner Program that takes advantage of the STM32 hardware Image Signal Processor (ISP) to offer an interface that helps streamline workflow by walking users through tuning steps more organically, showing the impact of settings in real time, and even using AI to automatically tune parameters. Emmetra showed how AUTOIQ.ai could improve the quality of an image sensor on an STM32N6 and an STM32MP257 using desktop software running on Windows and Linux. A macOS version should arrive next year. AUTOIQ.ai demystifies the image pipeline by handling some of the most technical and complex aspects so developers can focus on getting the best data.

What are some of the most common challenges ISP tuning software faces? #

Satisfying a broad range of engineering experiences

When ST released STM32 ISP IQTune, we announced that the software could “help developers tune their image signal processor (ISP) for a particular CMOS image sensor and its lens without needing the services of a third party or extensive expertise.” This continues to be true for many in our community. However, some teams require more handholding, especially when they have little experience with image pipelines. For instance, some decision-makers may underestimate that ISP tuning happens in stages and therefore demands more time and iterations than initially planned. Others may have a specific application in mind, like AI vision, and thus need assistance with key indicators than a general-purpose tool like STM32 ISP IQTune can provide.

Helping choose the right sensor

Another challenge is choosing the sensor itself, which goes far beyond selecting a resolution. There’s one common thread, regardless of the CMOS device in hand. Between the sensor’s MTF, dynamic range, or pixel architecture, choosing the right device can take a lot of time and resources. And settling on the wrong component can have dire consequences. One way to solve this is to use a repeatable, standardized workflow that helps quickly tune each sensor under evaluation. However, that requires a strong grasp of the underlying architecture, including sensor drivers, calibration, and pipeline tuning, among other things.

Supporting very different computing architectures

Implementation can also be complex, especially when working with different architectures, mainly a microcontroller and a microprocessor. Most already know that configuring a sensor for an STM32N6 running a real-time operating system is quite different from a profile on an embedded Linux machine using an STM32MP2, for instance. Unfortunately, knowing about it is not the same as being able to work out the intricacies of both approaches, which can require significant skills and experience. Many companies run computer vision applications on both an MCU and an MPU to support a wider range of use cases, but can’t always invest in the resources needed to become experts in both implementations. This is where a solution like AUTOIQ.ai stands out.

How does AUTOIQ.ai transform ISP tuning? #

Reducing friction

Customer adoption of AUTOIQ.ai is one of the strongest testimonies to the positive reception behind this software. Put simply, our customers shared that it reduced friction and helped them finalize designs faster, leading to new products that would have required significantly more time and investment. That’s also why ST works closely with Emmetra to ensure AUTOIQ.ai can take full advantage of the STM32 ISP middleware, as well as our software packages like X-LINUX-ISP and X-CUBE-ISP, so their solution remains compatible even as we release major updates, guaranteeing consistency with the entire STM32 ecosystem.

Outsourcing AI to AI

Beyond simplifying traditional ISP tuning, AUTOIQ.ai introduces a unique AI-centric tuning approach designed to maximize the performance of AI vision applications. While conventional ISP tools optimize image quality for the human eye, AUTOIQ.ai will tune it for the neural network itself. Users simply import their model, define the relevant KPIs, such as object detection confidence level, and AUTOIQ.ai automatically analyzes the algorithm and fine-tunes ISP parameters to improve AI performance. This enables significant gains in challenging conditions, including high-contrast or difficult lighting environments, without modifying the AI model or spending tremendous time in trial and error. By directly linking ISP optimization to AI outcomes, AUTOIQ.ai delivers a unique capability that many are using as a new benchmark.

Delegating complexity

AUTOIQ.ai has been successful because it offers a unique solution to the engineering challenges we mentioned above. Under the hood, the software tackles three major aspects of development, among other things: drivers, calibration, and pipeline tuning. By working closely with ST, Emmetra can optimize low-level code to extract features from sensors and improve performance. This enables more advanced calibration capabilities that help teams use cost-effective lenses or sensors, since they can get more out of them. Finally, because image pipelines are highly intricate and intertwined, AUTOIQ.ai handles this complexity so users don’t have to worry about how one change can affect other ISP blocks, among other things.

How to get started? #

The best way to get started is to grab an STM32N6 Discovery Kit or an STM32MP2 board with a camera module and request a demo from Emmetra. The company has already showcased at ST booths how its software can help tune a sensor’s ISP in hours rather than months. Concretely, users import RAW image files and see how the software automatically adjusts settings while still letting them tweak numerous aspects of the process to improve results. AUTOIQ.ai also offers a strong preview mode to see the difference before and after applying a setting, and it can run color analyses, fix exposure errors, and improve white balance or the signal-to-noise ratio.

── more in #computer-vision 4 stories · sorted by recency
── more on @emmetra 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/autoiq-ai-ai-powered…] indexed:0 read:5min 2026-11-17 · —