Luxonis turns six years old this week! In this update we will share details of the recent progress that we’ve made along with outlining what you can expect in the future. 

To start with, it’s an absolute privilege to get to work with some of the brightest and most driven engineers on the planet to help our customers win. The most rewarding part of the job is getting to see our customers reach scaled commercial deployments with our technology. Our customer’s use cases vary across a variety of robotics and automation applications. When we replace legacy technology or pioneer new ways of solving problems, we are helping to save lives and deliver productivity gains. The metric that we obsess about the most is our business conversion rate, which represents the portion of our customers reaching commercial scale versus those that are only running pilots. We have seen this metric increase 5x in the last four years! 

Our team has continued to grow as we mature our complicated technological stack and use it to tackle the ever-changing complexities of the physical world. Half of global GDP is driven by physical human labor that requires the ability to see and perceive the world in real time. At Luxonis, our mission is to enable systems that can exceed humans at every possible task or project in the physical world. We are building essentially the best perceiving superhero on the planet; one that can see, sense, and process the physical world better than any human can. 

This is why our devices provide sensor fusion on steroids with high resolution color cameras, stereo vision, Infrared, LiDAR, Thermal, IMUs, and microphones all synced together in unison dozens of times per second. And this is why our engineering team now consists of 80+ professionals who believe in our mission and drive rapid, iterative improvements across a variety of disciplines including electrical, mechanical, industrial design, firmware, cloud, applications, machine learning, synthetic data, and computer vision. We are always looking for great talent, and if you are interested in joining our team, please reach out to [careers@luxonis.com](mailto:careers@luxonis.com). 

Recently, we’ve released a number of new and exciting products. In the new sensing category, we have introduced new products including our first Thermal device (here) and our first Time of Flight device (here). To enable the ability to see and perceive depth at far distances, we’ve launched our Long Range device (here) with a 15cm stereo baseline that allows for accurate depth at greater than 30 meters. And most exciting of all, we’ve continued to roll out and scale up manufacturing of our new OAK Series 4 of products. The OAK 4 has everything that you love with our current OAK lineup, along with some key new features: 

• Power & Connectivity available with either USB or Ethernet 

• Standard 48MP RGB camera with support available for up to 108MP 

• An Octacore ARM processor running YOCTO Linux 

• 48 TOPS for AI Inference 

Series 4 is an edge AI sensor fusion powerhouse allowing you to do more on the edge in an all-in-one device than you’ve ever been able to do before. OAK 4 is your new go-to to option if you require a standalone device for automation use cases, along with if you need more capable sensors and processing in your larger robotics or automation system. A quick word on where this leaves our original OAK Series 2 lineup  

Our Myriad X based based OAK Series 2 is here to stay and remains the best option for a variety of use cases. If you need a stereo camera as a peripheral device with industry leading stereo (here) then OAK Series 2 is an excellent option. Luxonis has a large amount of customers who have spec’d in the OAK Series 2 for their robotics or automation applications and plan to use it for years to come. This means that Luxonis customers can expect 1) Reliable supply of the OAK Series 2 for years to come (here), 2) Continued rapid pace of new DepthAI releases every few weeks, and 3) High touch support for helping OAK Series 2 solve your problem at scale. 

If you want to give us input on what we should improve and what we should do next with our products, please reach out. 

The last exciting thing that we’ll share is that Luxonis is soon to be introducing a patent-pending technology that allows our stereo depth to be self-healing. Available now upon request and planned to be integrated into DepthAI by early this Summer, this new technology gives our depth algorithm the ability to dynamically recalibrate in the field. This means that you can expect to have a vision system that always works: it works out of the box and it works even after it’s gone through mechanical & thermal stress.

Here are some additional rapid-fire items that have happened in the last year at Luxonis: 

• We have scaled up manufacturing at a new location in Taiwan 

• We opened an office in Austin, TX bringing our global office total to six 

• We revamped our website:  https://www.luxonis.com/ 

• We provided only our second pricing refresh in the last six years, you can still place orders at current prices through April 20th, 2025:  here

For those of you who have lived startup life, you’ll know that it is a roller coaster with lots of extreme ups and downs. Fortunately, we love it and know that it’s all a part of the journey in bringing our mission and engineering efficiency to our customers. We believe that our 7th year in business is going to be our best yet. Stay tuned! 

@ThomasA We will have OAK ToF units available again in approximately 6 weeks. Thanks for your patience! The sooner you place an order the higher up on the list for shipping you'll be.

@erik Have you worked on anything for this?

@Luxonis-Adam Can you share when the V3 version of DepthAI core library will be released?

@LianyuChu You can use our guide here to get started: https://docs.luxonis.com/hardware/platform/deploy/poe-deployment-guide/

Yes, that's correct, +$200 versus RVC2 POE.

Good question, you can expect RVC4 units to be priced at approximately +$200 versus RVC2.

Luxonis recently celebrated our 5th birthday of being in business and we’ve never been more optimistic about what the next 5 years will bring. In this blog we’ll review our founding story, major milestones, growth, our product evolution, and our future plans. 

Founding Story 

Our journey started back in 2019 when we were inspired to save lives by making bike commuting safe. Sadly, the lives of countless cyclists are lost, and many more are injured due to distracted drivers colliding with them. We set out to solve this problem by leveraging computer vision and machine learning to be able to automatically detect when a driver is on a collision course with a cyclist. Check out the original rendering for our first product idea named the Commute Guardian.    

When we set out to create the Commute Guardian, we discovered that creating it with off the shelf components was extremely difficult. We needed to combine together multiple disparate pieces of hardware such as a stereo camera, processor, and AI accelerator. Finding the hardware was actually the easy part, the hard part was figuring out how to get the firmware to work. 

Milestones 

Discovering these challenges, we figured that we weren’t the only ones that were finding it extremely time consuming and expensive to put together a single computer vision system. This key insight led to the launch of our first products which were the OAK-D and OAK-1 that we introduced with a Kickstarter campaign. The OAK lineup delivered on five key elements that our customers required:

• Embedded:  Small size, low weight, low power, fast boot 

• Performant:  High resolution, high frame rate, low latency 

• Spatial:  Disparity Depth 

• AI:  Neural inference, object detection, semantic segmentation

• CV:  Feature extraction, motion estimation, edge detection, optical flow, WARP and de-WARP

With $1.4M raised across more than six thousand backers, we broke Kickstarter records for our product category.

We quickly followed the success of our OAK-D Kickstarter campaign with an even more affordable OAK-1 Lite and OAK-D Lite which Luxonis continues to offer at an entry price of $99 and $149, respectively. We launched our Lite versions with a second Kickstarter campaign and our customers were delighted with our efforts to democratize access to AI computer vision devices.  You can see a timeline of our key milestones below:  

Growth 

With the launch of OAK-D and OAK-D Lite, Luxonis began scaling up our manufacturing and global distribution efforts. We have shipped products into over 110 countries around the world and now hold inventory locally in our key markets which include North America, Europe, and Hong Kong for serving the APAC region. 

To support our customers' growing needs, Luxonis also expanded our global team which now includes 80 engineers. Luxonis performs all engineering disciplines in house and these include: 

• Electrical

• Mechanical

• Industrial Design 

• Firmware

• Software 

• Computer Vision 

• Application development 

• Machine Learning 

• Simulation 

With over 1 million SDK downloads, Luxonis now serves more than 25,000 customers around the world. We pride ourselves with our customer support and that includes our extensive documentation and our technical forum. Drawing from these resources, Luxonis recently introduced a ChatGPT-4 powered technical support chatbot that is now fielding hundreds of queries each week. 

You can learn more about examples of the various industrial applications that Luxonis works on here.  

Product Evolution 

We are focused on executing our product strategy which is highlighted by three key elements: 

1. Create a new generation of hardware that is 10x better every 2 years. Luxonis will be achieving this with our upcoming launch of Series 4  later in 2024. The pace of progress in edge AI hardware is remarkable and Luxonis will continue to be on the bleeding edge with Series 5 planned in 2026. 

2. Continually add new and improved sensing capabilities to our hardware. Luxonis recently introduced our first Time of Flight device that is available now for purchase and we are in the process of launching our first Thermal device that is available for pre-order now. We also continue to make improvements to our optics by offering a 32MP camera and we plan to introduce a 108MP option in our Series 4 lineup.  

3. Constantly move up the technology stack to make Luxonis a one-stop shop for robotic perception and AI automation. We have been scaling up our cloud and application Hub with a growing list of open-source, publicly available, and free applications including a Counter, QR code reader, and Object Detection.   

Our product philosophy is to both never end-of-life our products and to cannibalize our existing product with an always improving product offering. Our view is that if we don’t come out with new products that beat our existing ones then somebody else will. 

Looking Ahead 

At Luxonis, our ultimate goal is to enable systems that exceed humans at every possible task. Providing, at a minimum, parity with human-level perception means that robotic and automation systems are able to reliably assist, if not outright replace, various tasks that humans are performing today. We are inspired by achieving this goal and believe that the benefits to society will be massive. The robotic and automation systems that we enable help to save lives, eliminate grueling physical labor, and remove the need for humans to perform mundane repetitive tasks. 

Thank you to all of our stakeholders who have been along for the amazing ride over the last 5 years. We are incredibly grateful for all of the support and are extremely excited about what the next 5 years will bring. 

You can get in touch with me by emailing me at [bradley@luxonis.com](mailto:bradley@luxonis.com). 

Evolution gave humans, and most all other mammals, two eyes. Human eyes are typically spaced 6 to 7 centimeters (2.4 to 2.8 inches) apart. Remarkably, a substantial portion of the human brain is dedicated to processing visual information, with some studies suggesting a significant investment of the brain's resources in this area (source: University of Rochester). Our brain is able to perceive both how close an object is from us and where that object is heading. When Luxonis set out to create an edge artificial intelligence (AI) depth camera, we used learnings from human eye sight to design our products. This is why the stereo baselines of our cameras are typically 7.5 centimeters (3 inches) apart and a sizable portion of our onboard processing is dedicated towards fusing the left and right cameras together to create a three dimensional map. 

At Luxonis, we are on a long term mission of empowering companies to surpass human perception in every task. To achieve that mission, it requires the accuracy of the depth map to be superior to humans. Fortunately, we have already long since exceeded this because human beings are actually quite poor at estimating distances. If you look at your hands do you really know if they are currently 10, 12, or 14 inches away from your head? There is a reason we all have a tape measure (or laser or smartphone) at home to help us determine distances. 

We have recently been working diligently to improve the accuracy of our depth maps. And we are now proud to announce that starting in March 2024, stereo cameras shipped from Luxonis will be implemented with a next generation calibration procedure that delivers much better accuracy than our old generation. Please note that fortunately the performance is not hardware limited so existing devices can be recalibrated. 

Here are the real world (not theoretical) depth accuracy measurements that you’ll be experiencing with our products come March: 

• 0.7m - 4m (2’4” to 13’1”): below 2% absolute depth error

• 4m - 7m (13’1” to 23’): below 4% absolute depth error

• 7m - 12m (23’ to 39’4”): below 6% absolute depth error

You can learn more about our improved performance here:

https://docs.luxonis.com/projects/hardware/en/latest/pages/guides/depth_accuracy/#p-75mm-baseline-distance-oaks

If you’d like to get your hands on an OAK with improved depth sooner please reach out to us at [support@luxonis.com](mailto:support@luxonis.com).

We'll be releasing competitive benchmark data on depth performance by the end of March.

The error message you're seeing, RuntimeError: No available devices, indicates that the DepthAI library is unable to find your OAK D Lite device. Here are a few steps you can take to troubleshoot this issue:

1. Check your USB connection: Make sure your OAK D Lite is properly connected to your computer via a USB3 cable. If you're using a USB2 cable and want USB2 bandwidth, you need to specify the USB2 protocol. A bad USB cable can often cause connection issues.

2. Discover connected devices: You can use the DepthAI API to discover all connected OAK cameras. Here's a code snippet that finds all OAK cameras and prints their MxIDs (unique identifier) and their XLink state:

$$
import depthai
for device in depthai.Device.getAllAvailableDevices():
print(f"{device.getMxId()} {device.state}")
$$

3. Specify the device: If you have multiple OAK devices connected, you may need to specify which device you want to use in your code. You can do this by providing the MxID, IP address, or USB path to the depthai.Device constructor:

$$
device_info = depthai.DeviceInfo("14442C108144F1D000") # MXID
with depthai.Device(pipeline, device_info) as device:

...

$$

4. Check for device recognition: If you still can't connect to the OAK camera, you can check if your system recognizes the device by executing lsusb | grep 03e7 in your terminal. You should see a line similar to Bus 001 Device 120: ID 03e7:2485 Intel Movidius MyriadX.

If none of these steps resolve your issue, please provide additional information such as the name of your OAK camera, the version of the DepthAI library and bootloader you're using, and any crash logs or error messages you're seeing.

Sources:

• https://docs.luxonis.com/projects/api/en/latest/tutorials/multiple/#discovering-oak-cameras

• https://docs.luxonis.com/projects/api/en/latest/tutorials/multiple/#selecting-a-specific-depthai-device-to-be-used

• https://docs.luxonis.com/en/latest/pages/troubleshooting/#depthai-can-t-connect-to-an-oak-camera

• https://docs.luxonis.com/en/latest/pages/support/#depthai-issue

Hi Max, Luxonis can provide you with a refund if you can return these two units to us. Please email support@luxonis.com if you'd like to do that.

To deploy your custom grayscale Yolo8 model on the OAK device, you need to convert it into a MyriadX blob format. This process involves two steps:

1. Convert the model into OpenVINO's Intermediate Representation (IR) format using the Model Optimizer.

2. Compile the IR format into a blob file using the Model Compiler.

However, since you're encountering an error while converting the model to ONNX, it's possible that there's an unsupported layer or connection between two layers in your model. You can use the Netron app to visualize your model and identify any unsupported layers or connections.

If you're still encountering issues, you can try using the Blobconverter tool, which allows you to convert and compile the model from various formats, including TensorFlow, Caffe, ONNX, and OpenVINO IR.

Once you have the blob file, you can deploy it onto the MYRIAD-X processor within an OAK device for inference.

Sources:

• Luxonis Documentation: Conversion

• Luxonis Documentation: Model Conversion

• Luxonis Documentation: Troubleshooting

• Luxonis Documentation: Using Online Blobconverter

• Luxonis Documentation: Neural Network

• Luxonis Documentation: Local OpenVINO Model Conversion

To switch on/off the OAK-1 and OAK-D IoT 40 cameras to conserve power, you can use an external device like a microcontroller or a Single Board Computer (SBC) such as a Raspberry Pi. You can control the power supply to the cameras using the GPIO pins of the microcontroller or SBC. However, please note that the cameras do not have a built-in sleep mode or power-saving mode. They will start running as soon as they are powered on and will stop when the power is cut off.

As for customizing the timer remotely, you can use a device with network capabilities like a Raspberry Pi or an ESP32 board. You can write a program that runs on this device to control the cameras based on a timer. This program can be updated remotely if the device is connected to the internet. However, please note that the OAK cameras themselves do not have built-in network capabilities. They can only communicate with the device they are directly connected to.

For programming the cameras, you can use the DepthAI API which supports Python and C++. You can write a program that runs on the host device (like a Raspberry Pi or ESP32) to control the cameras. This program can be updated remotely if the host device is connected to the internet.

The IP65-rated OAK-D Pro W PoE camera is designed to withstand harsh environments, including dust and water jets from any direction. However, it's not rated for immersion in water or exposure to corrosive substances like seawater or harsh cleaning solutions.

The IP code of a device tells you whether it's dustproof and/or waterproof. An IP65 rating means the device is dust-tight and protected against water projected from a nozzle. It does not guarantee protection against immersion or corrosive substances.

If you need a camera that can withstand immersion in water, you might want to consider the IP67-rated PoE devices like the OAK-D-POE and OAK-1-POE. An IP67 rating means the device is dust-tight and can withstand water immersion between 15 centimeters and 1 meter in depth.

However, please note that even with an IP67 rating, the device may not be fully protected against corrosive substances like seawater or harsh cleaning solutions.

Also, when installing the camera, make sure that the GORE vent isn't blocked, as it can lead to moisture fogging up in front of the cameras (on the inside of the front cover).

TLDR: The only way to know is to actually test it out which is what we'd recommend that you try doing and we'd love to hear feedback on how well it holds up.

Yes, the connectors are indeed M series connectors from Amphenol. The M8 aux connector is an 8pin female A-coded connector, and the M12 ethernet is an 8pin female X-coded connector. The M12 plug does indeed go to a PoE cable, making this camera similar to existing PoE models. The M8 socket is used for auxiliary connections. If the M8 connector is not used, the cameras include an M8 connector cap for waterproofing.

To connect the camera to your computer, you will need a PoE switch or injector to power the device. After powering the device, you should connect your computer to the same Local Area Network (LAN) as the PoE device. If you are using a DHCP server, you can check the connected devices on the DHCP server dashboard to see if the camera is connected and its IP. If there is no DHCP, the camera will fallback to a static IP. Once you know the IP of the camera and can ping it, you can run any code sample or depthai experiment as you would when connecting an OAK device with a USB-C cable.

We are developing our Series 4 version of the Thermal camera and that new version will come with the option to connect via POE or USB.

The connectors on the back are M12 and M8.

The resolution is 256x192