AI Lounge

Voyager v1.7 landed and, past the usual new-models-and-hardware stuff, there are a few things that change what you can actually build on Metis that I thought it was worth us all taking two minutes to look at.The one that caught my eye right away is that re-identification is now a first-class task type. ReID is what lets you say "that's the same person who walked past the front camera two minutes ago," rather than treating every camera and every frame as a fresh unknown. If you've built a Home Assistant camera notifier and ended up hand-rolling logic to (try and) stop it pinging you about the same person fifteen times (which is something I’ve been battling with a lot personally), this is the release that lets you bin that code. There's a new tracking_with_classification.py demo that chains detection to tracking to per-track classification, which is basically the skeleton of a solid smart-security pipeline. Very cool.Depth estimation from a single camera is in there too, with a FastDepth

​Hi Everyone,​I’m Vishwanath Desai, a System Integrator from India. I am currently designing a smart surveillance and edge AI project named "Smart Dindori" for regional government schools and public infrastructure.​We are looking to leverage Axelera AI’s chips to power our multi-camera edge servers, and I would love to get some help from this community on two fronts:​Hardware Purchase: How and where can we officially buy Axelera AI chips/modules? What is the best route for procurement and delivery to India? ​Development/Coding Help: Is there anyone here who has experience working with Axelera’s SDK/software stack and would be interested in helping us with the coding and integration side?​If you are a developer open to collaboration or can point me to the right sales/technical contacts, please drop a comment or DM me.​Thank you in advance for your support!

Hi, I’m from India. Currently exploring AI and digital systems.Interested in learning AI automation, agent building, and real-world business use cases.Looking to connect and collaborate.

 Penalty shootouts in the Living Room ⚽We’re approaching World Cup season - I've been losing sleep to stressful penalty shootouts, and this little stress ball has been sitting on my desk the whole time. At some point the two facts collided: the most agonising moments of this tournament can come down to did the ball cross the line or not — the exact thing goal-line technology and VAR exist to settle. So I thought: could I build my own scrappy version of that and run it on the edge?In all fairness, this was really just an excuse to get my hands dirty with the Voyager SDK end-to-end - but a penalty shootout in my living room is a much better motivator than some benchmarks. 😅The match setup. The official match ball is the Axelera stress ball. The goalposts are… two plants in mugs. I roll the ball at the goal, and the system calls GOAL or MISS — and crucially, rebounds off the wall behind don't count, because the ball has to actually cross the line. I don't detect the goal itself; I just e

Hr department for BBA studys complete and after BBA for MBA start  

Voyager SDK v1.6.1 is a patch on top of v1.6.0. No huge new feature set, but useful cleanup, real Windows performance gains, closed-loop power control on the upcoming Metis M.2 Max, and a proper docs portal at docs.axelera.ai.Windows is FasterThat’s not something you hear about Windows very often, is it? But this is referring to Voyager on Windows, so that might explain it. 😋The PCIe driver and libaxldev have both had performance work, closing the gap with Linux, which the Windows fans out there will appreciate. Logging and tracing infrastructure is also improved, with a utility for viewing traces.By the way, before we go too much further, this release requires Metis Driver 1.3.2 or newer on Windows. v1.3.1 and earlier are no longer supported, so v1.6.1 ships with driver 1.3.11.Metis M.2 Max Power ControlClosed-loop power control on M.2 Max has been improved ahead of its launch:No more unexpected performance degradation. The limit set via axdevice --set-power-limit now matches what ax

I'm trying to deploy MitoNet v1 (a mitochondria segmentation model based on PanopticDeepLab) on the Axelera Metis AIPU using the Voyager SDK v1.5, and I'm running into issues exporting the model to ONNX.MitoNet is available via the empanada-napari library and its weights are stored as a TorchScript archive (.pth). When loaded, the model is a `torch.jit.RecursiveScriptModule` that takes a (1, 1, H, W) float32 tensor as input and returns a dict with keys `sem_logits`, `ctr_hmp`, and `offsets`.The problem is that the model uses `aten::__interpolate` with an `int[]` size argument, which blocks all ONNX export paths I've tried:1. `torch.onnx.export` with opset 11 and 17 → `INTERNAL ASSERT FAILED: We don't have an op for aten::__interpolate`2. `TS2EPConverter(model, args).convert()` → same `aten::__interpolate` error3. `axelera.compiler.quantize()` directly on the TorchScript model → `ValueError: Failed to get input shape from calibration dataset`Environment:- Voyager SDK v1.5- PyTorch 2.10.

I'm glad there's an AI hardware community. I'm a Python Flask app developer, and I'm currently creating an app for IoT remote automated control, filling materials into a large mixer. I'm currently having trouble getting the truck loaded with sand into the truck. I have to scan the material's position in the truck bed, how many cubic meters per truck, and the driver's data, along with the car's license plate. Can an AI chip help me solve my problem?

Hello everyone, my name is Mahameru from Indonesia. I'm currently conducting computer vision research using Axelera PCI and M2 cards. I've tried running the model using YOLO Model versions 8 and 11, and I get different confidence levels when running using the CPU/GPU (local on my laptop) and Axelera. Can anyone help me?

Heads up! Ultralytics is hosting a live session on 8th April all about running YOLO models on Metis hardware. Two of Axelera's engineers (​@A-Tang and ​@Steven Hunsche) are joining the Ultralytics team and Innowise to dig into how we're enabling high-throughput, energy-efficient inference at the edge.If you've been running YOLO models on Metis (or want to), this one's worth tuning in for. Expect a proper technical deep dive into what the Ultralytics + Axelera integration looks like in practice.It's Ultralytics Live Session 22, streaming on YouTube, or you can watch it right here in this post (below 👇).When: 8 April, 2026 at 15:00 CEST / 14:00 BST / 09:00 EST / 06:00 PDTWatch/register here: Ultralytics Live Session 22If you catch it live, watch it here, and drop any questions or takeaways in this thread afterwards. Always good to hear what lands. 👍 

Hey everyone,Excited to share a brand-new video I put together for anyone getting started with the Metis card and Voyager SDK. Whether you're setting up your eval kit or just curious about how inference works on Metis, this walkthrough is for you.🛠️ What’s inside:Setting up your environment and verifying camera input Running YOLOv5s-v7-COCO on a live camera stream Using download_prebuilt for faster setup Running inference on media files and datasets Benchmarking performance Measuring accuracy with mAP and more📺 You can find this tutorial, and more, on our YouTube ChannelThis guide is based on SDK version 1.2.5 and follows the official Quick Start Guide, but with some extra tips and tricks to get you up and running even faster.💬 Have you tried running inference on Metis yet?Drop your thoughts, questions, or your own setup experiences in the comments, I'd love to hear how it's going for you or help troubleshoot if you're stuck!See you there,Jonas 

👋 Hello Axelera AI Community,My name is Saravanan, and I’m an AI Architect and Executive Officer at Super Quantum AI, based in Gobi, Tamil Nadu, India.I focus on designing advanced AI architectures, multi-agent systems, and next-generation intelligent platforms. My work is centered on building scalable AI infrastructure, intelligent automation, and future-ready AI ecosystems.I’m excited to be part of this community to learn, share knowledge, and connect with fellow AI engineers, researchers, and innovators.You can learn more about me / my work here:• https://saravananai.online• https://eshmoai.onlineLooking forward to meaningful discussions and collaboration.Thank you!— Saravanan

Many of you guys have been on the ride since we opened up here (even before), so I thought I’d share a heads-up that Axelera has announced a new funding round of over $250 million! That takes total funding to over $450 million since the company was founded in 2021, and is apparently the largest investment ever in an EU AI semiconductor company. 😮The funding will go towards scaling manufacturing, expanding the partner network, and developing the SDK further, so good news for everyone here building on the platform! Thanks for being here with us!Check out more of the details here.

If you've wanted to peek under the hood of how your Metis hardware talks to Linux, there’s a new Axelera AI repo with the PCIe driver source code.Repo: https://github.com/axelera-ai-hub/axelera-driverFor most of us running the standard Voyager SDK setup, you won't likely need to touch this. The driver gets installed automatically. But if you're doing something a bit more adventurous (custom kernel builds, embedded systems, that sort of thing) this could be really useful. I know at least a few people on here have been experimenting with Debian and Yocto, for instance.Quick overview of what's there:It's the metis.ko kernel module. The one that handles all the PCIe communication between your system and the Metis hardware. Currently it supports Ubuntu 22.04 and 24.04 (LTS).Building it yourself is pretty straightforward if you should want to. Make sure you've got your kernel headers installed (sudo apt-get install linux-headers-$(uname -r)), then it's just the usual make and sudo make insta

Hello Axelera Community!My name is Ioannis and I am a part-time working student here. Before moving to the Netherlands to start my Master's studies, I was researching potential companies to intern with, and Axelera was a name that really stood out. When I attended a presentation by the company during my Master's kick-off events and heard they were looking for working students, it felt like the perfect opportunity!I officially joined in December and have been working with the Metis AIPU on the Lenovo P3 Tiny and an OrangePI since then. Over the last couple of months, I have worked on a variety of scenarios, starting from simply deploying models to building a mini demo that combines Whisper, LLaMA, and a text-to-speech model. I have also been documenting my experience along the way, testing the Voyager SDK documentation and evaluating how intuitive and user-friendly it feels for a newcomer.When it came to learning the new hardware and software, the guides were very helpful. However, I fo

Hi Axelera inferences and compiler available https://hub.degirum.com

Hello everyone,Abdi here. I am a part-time working student at Axelera. I joined in December.I am having a blast playing with the Metis devices. Our task is focused on testing the Voyager SDK documentation. Is it clear enough for the customer? Where can it be improved? We test the documentation by actually following it, such as working through the tutorials provided in the documentation.Furthermore, we test how easy it is to perform tasks that users may actually want to do, such as counting how many objects are in a video frame using YOLO. Deploying models,running different kinds of models.We evaluate how useful the documentation is in this context and identify where the gaps are.Currently, I am working on integrating an external tool, Whisper, with the Metis Dev System on the Arduino Portenta X8, and  routing Whisper’s output into an LLM running on the device. -Abdi   

I’m interested in hearing from everyone here about which models you’re successfully using with your Metis device and Voyager SDK, and which ones you’d like to see being officially supported.This could be really useful info in terms of prioritisation when it comes to adding more support in the model zoo.

Hi everyone!This is my first post in the community, so let me introduce myself! My name is Pablo and I’m new to Axelera, joined in December as a part-time working student.Throughout these last two months I’ve been using the Metis AIPU both in a Raspberry Pi 5 and with the Advantech MIC-770, running through several use cases and analyzing documentation and user experience. I started with the most basic setup tasks before making my way to all Voyager SDK’s tutorials and examples/demos, and even experimented with combining external tools like Whisper into the system’s LLMs. It’s indeed quite fun and impressive to see what Metis is capable of! (and having such interesting projects on the community clearly proves that 😄)By acting as a sort of customer that is playing with a new system, I hope that my perspective is relatable and helps Axelera to keep improving. I have to say I’m very happy with how my experience is going so far, and can’t help but feel more excited about edge AI every day!

Hi Axelera community,I’ve just started working with the Metis SBC and Voyager SDK, and I’m looking for some structured guidance on how to get up to speed.My background:~10 years of hardware experience (board-level design, complex FPGA design using VHDL/Verilog) Comfortable with Python and have worked with AI models at an academic level Limited embedded experience beyond bare-metal concepts (I’ve usually worked alongside dedicated embedded engineers)I’d love suggestions on: A learning path for Metis + Voyager SDK Where to start (simple “hello world”–style projects) How to progress toward more complex designs (e.g., integrating peripherals, using accelerators, optimizing performance, etc.) Career/skills development advice For those who have followed a similar path, from low-level hardware/FPGA work into higher-level embedded Linux and SoC-based systems: How did you structure your learning? What skills or topics were most valuable (e.g., Linux drivers, build systems, device trees, per