Your video is the best tutorial Ive ever seen. Focused, uniform flow (no time wasted, like watching someone tighten 4 nuts for 90 seconds). Perfectly clear and understandable speaker, no excessive dwelling on sales/products. Was this a professionally managed video (scripted/rehearsals etc) cause it always moved forward. Great job.
got my AI Kit today - need to start working out on identifying the objects and creating some models.. Lovely presentation with clear details and explanation. Love how it is presented.
Currently LLMs are not supported mainly due to the memory bottleneck on the PCIe. LLMs requires a huge amount of parameters. And the performance will be very limited due to memory access.
This is amazing, I'll be getting one for my rpi5. I want to build a Lego sorter that uses camera vision to identify and sort my son's Lego. I brought a pixycam2 in hopes of an easy short cut but it can only store 7 pesets. Also didn't get an library update when the r4 dropped, not yet anyway (I'm hopeful) Does this product have shape detection? I'm not a software engineer so having libraries or "pipelines" with features to assist in the build would be great help. Lego do a thing with rpi3 so they would probably cross brand with you on an official Lego sorter set. Thats if your looking for more publicity on this product. Thank you for this tutorial and giving makers and hobbyists a cool pi hat for home builds.
i am trying to work out lego sorter - to find pieces when a picture of it is shown to the eye... couldnt succeed yet.. can u share your project ? got my AI Kit today - need to start working out on identifying the objects and creating some models..
Very detailed and easy to understand video and a cool product!. Thank you. 1) Will the model recognize a train model on the go? Are you planning to do a video about custom object detection?
great tutorial! I'm excited to experiment with my new Pi 5 + AI Kit. Pls make a video tutorial showing step by step how to use a Jupyter Notebook to fine tune and deploy a YOLO model on the RPi AI Kit. I know you've got written documentation but a video is always a welcome addition : )
Hope we get home assistant integration fingers crossed I just got one to see if I could try to figure it out as a beginner this video helps to get me in the ball park looking forward to frigate integration as it’s also a home assistant integration very exciting stuff
Your Model Zoo is cool - but are there any tutorials - how to design-compile-train-deploy a quantized CNN model that will run on your 8l TPU? Or transfer-learning retrain of a pre-trained CNN model at least? Any links will be appreciated.
Can you make a video detailing Hailo8 installed into a Mini PC or PC/server x86 pure Debian platform for a full home Frigate implementation? This is what I'm currently working on. I am currently using Coral TPUs but want to buy Hailo M.2 modules without the Pi hat. Thank you!
I got one of these, and it works perfectly with the provided sample apps. However, it has a steep learning curve to start coding something on a Pi to leverage the Hailo kit. It is not suitable for everyone using a Pi unless you are proficient with C++.
Hi checkout Hailo basic pipelines in our repo. All in python with a very simple learning curve. Would you like a tutorial for the basic python pipelines?
@@giladnahor Thanks for your reply. A tutorial would be nice. I did try to use those Python scripts but I ended up with a lot of errors. May be having a tutorial would be easier to understand how to use it. Thank you for your help
Hailo does support a full SDK for importing networks from multiple frameworks. Onnx, Tensor flow, pytorch and more. In these examples we aim for simple pre-built pipelines.
@@giladnahor Can you make video tutorial for work with Mvtech Halcon Software like this video ? . Now I stuck cann't install and use it on windows and Linux.
H,i raspberry pi does not support Ubuntu officially. Therefore Hailo cannot support it. There is a topic in the community from a guy that managed to get it up and running. But it requires a lot of manual installations....
[Help] Hello, I'm trying to convert a custom YOLOv8n ONNX model to HEF format on Windows. I've tried several methods but haven't succeeded: 1. Using hailo_ai_sw_suite_2024-10 docker container, I tried: - Direct command line conversion: ```bash hailo parser onnx --input-files /input/best.onnx --output-har /output/model.har hailo optimize --input /output/model.har --output /output/optimized.har --use-random-calib-set hailo compiler --input /output/optimized.har --output /output/ice_pack_detector.hef 2.Using Model Zoo approach: hailomz parse yolov8n hailomz optimize yolov8n hailomz compile yolov8n --hw-arch hailo8 Using DFC Studio web interface: Ran docker with port mapping: docker run -it --rm -p 3000:3000 -v "C:\hailo_temp:/input" -v "C:\hailo_temp:/output" hailo_ai_sw_suite_2024-10:1 Accessed localhost:3000 Created new project but encountered parsing errors with Start/End node configuration I've reviewed the official documentation (which is quite brief) but still can't successfully convert my ONNX model to HEF. Questions: Is there a simpler way to convert ONNX to HEF on Windows? What's the recommended approach for Windows users? Are there any specific requirements or configurations I'm missing? Environment: Windows 10 Docker Desktop YOLOv8n custom model (ONNX format) Hailo AI Suite 2024-10 docker image
Script kiddie but coding noob (suck at writing but ok with reading, stealing… I mean copying and modifying code) here 😔 How would I go about or where would I even start if I wanted to input a video feed of an office lounge/café area to track and report on the utilisation of each seating group/area to be able to judge how well certain seating arrangements/layouts work for and are accepted/used by staff? Would be awesome to get stats on: people standing/passing through the area vs the ones who sit down for let’s say longer than 30 seconds. And, because “it’s so easy”, why not also have stats on how long the average stay is, whether it’s a single person or two or more people interacting, whether they’re reading, working on a laptop or maybe taking a call. This might tell us if we… have enough space(s), whether they’re the right type (e.g. for groups vs tuck away/quiet-booths), if window with park view or the more distraction free ones are used more, … 🤓
Good luck trying to buy anything associated with a Pi. I've been trying to build a cm4-IO-board for over two years. Out of stock. Out of stock. Out of stock. No Hailo. Out of stock. Hailo web site also refuses to take orders from people with gmail or outlook. No stock of hailo chip on Pi Hut. Out of stock. Out of stock. Out of stock.
Your video is the best tutorial Ive ever seen. Focused, uniform flow (no time wasted, like watching someone tighten 4 nuts for 90 seconds). Perfectly clear and understandable speaker, no excessive dwelling on sales/products. Was this a professionally managed video (scripted/rehearsals etc) cause it always moved forward. Great job.
got my AI Kit today - need to start working out on identifying the objects and creating some models.. Lovely presentation with clear details and explanation. Love how it is presented.
Can we accelerate ollama models. We want LLM acceleration.
Currently LLMs are not supported mainly due to the memory bottleneck on the PCIe. LLMs requires a huge amount of parameters. And the performance will be very limited due to memory access.
Their 10H will solve that. Question when it will be released …@Hailo
love to see it working with network camera aswell as Frigate integration
This is amazing, I'll be getting one for my rpi5. I want to build a Lego sorter that uses camera vision to identify and sort my son's Lego. I brought a pixycam2 in hopes of an easy short cut but it can only store 7 pesets. Also didn't get an library update when the r4 dropped, not yet anyway (I'm hopeful) Does this product have shape detection? I'm not a software engineer so having libraries or "pipelines" with features to assist in the build would be great help. Lego do a thing with rpi3 so they would probably cross brand with you on an official Lego sorter set. Thats if your looking for more publicity on this product. Thank you for this tutorial and giving makers and hobbyists a cool pi hat for home builds.
i am trying to work out lego sorter - to find pieces when a picture of it is shown to the eye... couldnt succeed yet.. can u share your project ?
got my AI Kit today - need to start working out on identifying the objects and creating some models..
Thanks 🙏 Just set up my pi 5 today 🎉
This exactly what I been waiting for and I'm really looking forward to try this out 😃
Very detailed and easy to understand video and a cool product!. Thank you. 1) Will the model recognize a train model on the go? Are you planning to do a video about custom object detection?
I am looking forward to a video on Frigate integration.
great tutorial! I'm excited to experiment with my new Pi 5 + AI Kit.
Pls make a video tutorial showing step by step how to use a Jupyter Notebook to fine tune and deploy a YOLO model on the RPi AI Kit. I know you've got written documentation but a video is always a welcome addition : )
Do you make Ollama faster? Or any local LLM?
Hope we get home assistant integration fingers crossed I just got one to see if I could try to figure it out as a beginner this video helps to get me in the ball park looking forward to frigate integration as it’s also a home assistant integration very exciting stuff
Excellent video.
How about a license plate detection project any hope on that?
Can I use an RTSP camera for video feed instead of a USB or RPI camera?
love to see it working with network camera and have Frigate integration
Custom model yolov8 with python api.
I'd like to see a demo of AI kit recognizing hand gestures
hello what command do i use to see the performance of the Hailo module? Like the one found at 4:56 that shows memory usage
Next time please don't use music background 🙏 very well documented 👍
Got mine.
Working great so far. But what else can I do with it?
That's what you need to figure out.
Why not write all commands in one nix file ? to give possibility to install with one command, and be easy to explain
Your Model Zoo is cool - but are there any tutorials - how to design-compile-train-deploy a quantized CNN model that will run on your 8l TPU? Or transfer-learning retrain of a pre-trained CNN model at least? Any links will be appreciated.
If Hailo want to sell a tons of these, please make more videos and instructions of use case.
Can you make a video detailing Hailo8 installed into a Mini PC or PC/server x86 pure Debian platform for a full home Frigate implementation?
This is what I'm currently working on. I am currently using Coral TPUs but want to buy Hailo M.2 modules without the Pi hat.
Thank you!
can this be run on a pi5 running ubuntu? is there documentation for this?
oh yes frigate integration would be awesome!
Will the AI kit work on Raspberry PI 4 ?
I couldn’t use this anyway because it uses a port that I already use
Is it possible to use a few (3 or 6) of these at an event space to capture the number of visitors at an event (like a trade fair or open house)?
Why do you use Display cable for camera ?
You are right, this is by mistake 😅
I got one of these, and it works perfectly with the provided sample apps. However, it has a steep learning curve to start coding something on a Pi to leverage the Hailo kit. It is not suitable for everyone using a Pi unless you are proficient with C++.
Hi checkout Hailo basic pipelines in our repo. All in python with a very simple learning curve. Would you like a tutorial for the basic python pipelines?
@@giladnahor Thanks for your reply. A tutorial would be nice. I did try to use those Python scripts but I ended up with a lot of errors. May be having a tutorial would be easier to understand how to use it. Thank you for your help
Connect it with Immich for self hosted photo library classification!
That's a great idea I will review it with the CLIP network. This will allow for free text search.
Can this accelerator run games with ray tracing on ?
9:16 errors/inconsistent, TAAPAS_POST_PROC_DIR or TAPPAS, // in directory name works but ...
13:09 defult => default (hailo app help)
Does it support various machine learning libraries/frameworks?
Hailo does support a full SDK for importing networks from multiple frameworks. Onnx, Tensor flow, pytorch and more. In these examples we aim for simple pre-built pipelines.
@@giladnahor Can you make video tutorial for work with Mvtech Halcon Software like this video ? . Now I stuck cann't install and use it on windows and Linux.
Great demo. Where can I order it in Israel?
Pitel. Go to raspberry pi site and check out official resellers in Israel.
i have rasberry pi 5 with hailo but i boot Ubuntu. Can i get hailo driver on this device? or i must reinstal tu rasbery pi OS ?
H,i raspberry pi does not support Ubuntu officially. Therefore Hailo cannot support it. There is a topic in the community from a guy that managed to get it up and running. But it requires a lot of manual installations....
bummer 😢
Wait, does the display cable work with the camera module? Cause it didn’t work for me when I mistakingly tried it.
Pi 5 needs a new cable :(
YEA BOY
Am getting Bus error, anyone know how to solve it
Would have been good to see it detecting the all the cars and vehicles outside your window live.
Have you driver for Ubuntu ? I have o nmy rasberi pi ubuntu
and your NPU
will the 8L work on x64 machines?
Yes it will
Where is Hailo from?
I think it is from Israel.
Oh hell no. I'm selling my AI kit then.
@@okra3000 🤣
[Help]
Hello, I'm trying to convert a custom YOLOv8n ONNX model to HEF format on Windows. I've tried several methods but haven't succeeded:
1. Using hailo_ai_sw_suite_2024-10 docker container, I tried:
- Direct command line conversion:
```bash
hailo parser onnx --input-files /input/best.onnx --output-har /output/model.har
hailo optimize --input /output/model.har --output /output/optimized.har --use-random-calib-set
hailo compiler --input /output/optimized.har --output /output/ice_pack_detector.hef
2.Using Model Zoo approach:
hailomz parse yolov8n
hailomz optimize yolov8n
hailomz compile yolov8n --hw-arch hailo8
Using DFC Studio web interface:
Ran docker with port mapping: docker run -it --rm -p 3000:3000 -v "C:\hailo_temp:/input" -v "C:\hailo_temp:/output" hailo_ai_sw_suite_2024-10:1
Accessed localhost:3000
Created new project but encountered parsing errors with Start/End node configuration
I've reviewed the official documentation (which is quite brief) but still can't successfully convert my ONNX model to HEF.
Questions:
Is there a simpler way to convert ONNX to HEF on Windows?
What's the recommended approach for Windows users?
Are there any specific requirements or configurations I'm missing?
Environment:
Windows 10
Docker Desktop
YOLOv8n custom model (ONNX format)
Hailo AI Suite 2024-10 docker image
I want to buy your hailo v10 please. I so badly want a pi 5 llama 8 b server!
Script kiddie but coding noob (suck at writing but ok with reading, stealing… I mean copying and modifying code) here 😔 How would I go about or where would I even start if I wanted to input a video feed of an office lounge/café area to track and report on the utilisation of each seating group/area to be able to judge how well certain seating arrangements/layouts work for and are accepted/used by staff? Would be awesome to get stats on: people standing/passing through the area vs the ones who sit down for let’s say longer than 30 seconds. And, because “it’s so easy”, why not also have stats on how long the average stay is, whether it’s a single person or two or more people interacting, whether they’re reading, working on a laptop or maybe taking a call. This might tell us if we… have enough space(s), whether they’re the right type (e.g. for groups vs tuck away/quiet-booths), if window with park view or the more distraction free ones are used more, … 🤓
STUPID
BACKGROUND
MUSIC
Will pull the github repo and read it.
Good luck trying to buy anything associated with a Pi. I've been trying to build a cm4-IO-board for over two years. Out of stock. Out of stock. Out of stock. No Hailo. Out of stock. Hailo web site also refuses to take orders from people with gmail or outlook. No stock of hailo chip on Pi Hut. Out of stock. Out of stock. Out of stock.
Found the boomer Trump supporter
robot ++
Please hire someone from airtasker ! To improve videos and ! It was distracting to watch with a stupid background music