DEV Community: BrainyPi

The story of Brainy pi - Know it all

BrainyPi — Fri, 07 Apr 2023 09:23:11 +0000

Brainy Pi is a single board computer made by entrepreneurs for entrepreneurs and businesses .
This blog will tell you why brainy pi exists ? The unsolved problem, the innovative approach , why us , comparision with peers and the future of Brainy pi. You will know it all at the end of it.

Why Brainy pi exists?
The Unsolved Problem: A lot of business ideas can do better with customised hardware .Compared to software hardware design still remains hard.Building custom software by integrating different modules has become common and easy which has brought so many niche software solutions .Can we get this modularity available in Software to hardware design and help push more next generation AI and IoT products faster into market? Raspberry pi and Arduino showed the world atleast faster prototypes are now surely possible .
But from Prototype to Mass Production it still remains a hard design journey and many kickstarter products never get delivered . So many creative ideas lost for humanity.

Image source : https://qualityinspection.org

Solution : Can we get the best of 2 worlds ? Speed while prototyping , Scalable design for Manufacturing ?
Brainy pi was created with this exact thing in mind. Brainy Pi builds further on the once great idea of Raspberry pi made primarily for prototypers.

The new approach: Brainy Pi is a full hardware and software solution that bridges the gap between prototyping and mass production. It leverages the software ecosystem of Raspberry pi for fast prototyping, but is customizable for mass production.

How about the code ? Wont we have to re do all the code if hardware is customised or changed?

The magic : Brainy pi backed by Shunya os team can ensure all code written for Brainy pi / Raspberry pi can be ported smoothly into any Arm(Cortex -A) based hardware opening a wide range of SoC(chip) choices in terms of configuration and pricing for their final product.

Q1) Really!! or is Brainy Pi just another clone of Raspberry Pi ?What is different ?
Every other clone of Raspberry pi is mostly a hardware clone trying to make a quick buck on the demand generated by Raspberry pi .
Brainy pi is not a clone but inspired by what they did .Brainy pi is a full hardware and software solution unlike others we don't plan to release too many boards .
Brainy pi will provide enterprise grade stability and support the software stack to keep it bug free , so people can easily build quality solutions on top of it .

Q2) Tell me more , a purpose , a goal that is different from other boards?
In one line : Stable , Reliable , Long term availability , well documented and supported , focussed on serious product developers and entrepreneurs .
Brainy pi has a clear reason to exist : Create an easy path for entrepreneurs/enterprises to go from existing/new prototypes to mass production . Brainy pi is designed with this in mind in both software and hardware.
The purpose is not to launch and sell a plethora of boards but to make a stable reliable board and maintain it long term as a platform of choice to build great products on .

Q3) I see many similar boards from Europe ,Taiwan and Chinese vendors? Can you compare a little more ?
The core chip RK3399 is used by many from Google Chromebook , Asus Tinkerboard 2, many Chinese vendors Orange , Nano , Rock and other pi ’s . All of them use Rockchip’s tablet pc reference design as the base and then modify as per their goals . So in hardware Brainy pi maybe anywhere between 90–95% close to all these boards since they all originate from the base reference design .
What we have focussed while designing the hardware is choosing components which will have low supply chain disruption .Components were substituted (e.g we sacrificed BT 5.0 since it will remain under pressure ) with ones which are more likely to not go through disruption . The board is designed with a purpose of being long term available .
In software too, we stick to our purpose, being easy yet stable and enterprise grade . Rbian(Raspbian compatible debian) is created with the idea of being compatible with Raspberry pi code .One of the key goals is that all Raspberry pi code should directly run on Brainy pi for ease of use still we use only LTS Linux kernels for reliability of system.
Further the team offers an option to use Yocto based distro(Shunya) which is custom , minimalistic and can be tailored exactly for the application needs when the prototype has to go into mass production and become more efficient .
We are the only team which has deep focus on software we dont stop at providing a compatible OS or porting raspberry pi libraries we offer AI and IoT application boiler plates , that can be used as a base to build real life solutions .

A Market wise comparision with vendors from other region :

Europe Vendors: Quality boards but higher priced support
US vendors : Quality board but hard to access support e.g Qualcomm high price not open to reveal chip design to everyone with sub 10k volume.
Taiwan : E.g Asus Tinkerboard 2 , Medium price but almost no support still offers older kernel 4.4 with limited support (mostly only android)
All Chinese vendors : Close copies of each other with low stability in both hardware and software Chinese Strategy is largely to release boards with more and more attractive paper features at lower prices , but not bother to stabilise software and offer accountability on released feature stability. Brainy Pi : Low price with good support . Strategy : Offer one stable board keep fixing bugs as they get discovered and maintain it for customers so they have an easy time designing products and dont have to deal with surprises . And yes , we come from India 🇮🇳 .Indians have learnt processes from west and now implements it at same standards at much lower price . The founder also lived in China 🇨🇳 for 8 years and understands hardware design and its supply chain very well . So in a way this is best of 2 worlds in hardware and software.

Q4)You said the hardware is designed to improve availability yet it is not available for instant purchase why do I have to apply ? Why cant you mass produce it and make it available for buying for all ?

Two reasons :

1) Funding : We had budget just enough to produce a minimum batch of 3000 (minimum the factory could accept for SMT ) By giving these 3000 pcs to the right customers(influencers and volume buyers) we need to find enough orders to keep producing atleast 5000 pcs every month. If we weren’t a bootstrapped startup we would do exactly that produce a lot and make it available to all .So if you place a large orders availability is solved by better design ,but it is only the money that stops us not the chips.
2) Quality support over quantity sales : We could have used the initial excitement made a quick buck selling all 3k pieces to just anyone who wanted to buy .We are in this for long term . Just short term selling would mean supporting all kinds of uses cases by prototypers , experimenters and customers who would try random things and support experience would suffer .
Our key differentiation is we are for serious folks and we will support the product and this would have compromised it .So we choose to select customers ,go slow but support them well .

Q5) Everything has a downside , What are the downsides of slow but stable approach ?
Sure !! There is always a trade off .
We are focused on serving the enterprise market, and our goal is to provide a stable product rather than constantly chasing the excitement of new but unstable features. That’s why you won’t see a lot of consumer-grade boards with the newer configurations from us. There are others who serve that market better than us. Our focus is on long-term stability for enterprises, so you don’t have to keep redesigning. .
We are like the premium brand phone that always gets its features late but is very stable .
Sorry to the hackers we know you want all the fun with new boards but we lean a bit towards commercially stable products.

Q6) If it is about long term why not choose a newer SoC from Rockchip like Rk3566/68 etc ?
Short answer : Google ,Industrial and AI variant available for same SoC

Google Logo (no association with Brainy pi)

For long answer read on
Choosing RK3399 was for the same enterprise reasons, against other more faster SoCs which are yet to be proven. Rk3999 is used in Google Chromebook and Google continues to work on its kernel since 2016 . It is updated at a very high frequency .We learnt and got a lot of fixes into our kernel from Google’s kernel repo .
Unlike Freescale or TI or Qualcomm Chinese SoCs BSPs just work but leave a lot of room for improvement .A customer like Google does lot more at software level. So we know there will be many unknown issues with the consumer grade SoC’s RK3566/68 etc and may remain unknown until a lot of it is deployed or an expert Company like Apple or Google looks through all of it .
Besides these consumer grade SoCs will have shorter lifetime so eventually by the time it gets stable the SoC may be out of supply. Stability over excitement in designs is what Brainy pi offers to its customers. Brainy pi founder was behind launch of many consumer products and has insider view of issues in SoCs which are not known to a new designer . Rest assured we offer something that is a great value to customer , cheaper* but not compromised.
*Compared to Asus Tinkerboard 2 ($179 list price) and Google Chromebook which uses this SoC we have kept our prices ex-factory in the range of 75$ .
Also , RK3399 is available pin to pin compatbile in two more variants Industrial grade temperature version and RK3399 pro (AI version) so the design will remain for long term for your growing application needs thats how we zeroed down on this one.

Q7) What are the future plans with Brainy pi?
We been on this since 2017 (as Shunyaos and IoTIoT team ).Hardware design started in Dec 2021. Currently while in beta we are focussed on 2 things, getting the board very stable in terms of software ,our early acess customers and testers report unknown bugs and we solve them promptly as we know it .
As next stage we plan to make it into a ready to use platform (think a ready made pizza bread ) on which customers just have to put their add ons(think toppings) , write code (think baking it) and validate in market . If they hit a home run with their prototype , we are always there to help them take it to next level .
And yes , the pizza bread thing is going to be way cool , expect a lot of AI and IoT projects and other useful ready product codes to be released soon on the website . Please check our documentation page and follow us on linkedin
Ps: There is an youtube channel which has many live projects with brainy pi. Check it out here.

Object detection using OpenCV on Brainy Pi

BrainyPi — Fri, 07 Apr 2023 09:10:40 +0000

Object detection is an exciting area of computer vision that has numerous practical applications. From self-driving cars to security systems, object detection can help machines make sense of the world around them. In recent years, deep learning-based approaches have shown great promise in achieving state-of-the-art performance in object detection tasks. However, implementing these approaches can be challenging due to the complex models and hardware requirements. Fortunately, with the advent of single-board computers like the Brainy Pi, it is now possible to deploy sophisticated object detection models using open-source libraries like OpenCV. In this blog post, we will explore how to perform object detection using OpenCV on the Brainy Pi, and provide a step-by-step guide to get started with this exciting technology.
Overview

Before diving into the code, let’s first discuss the project structure and the different files and folders used.The objectDetection.py file contains the Python code for object detection using the Single Shot Detector (SSD) algorithm with MobileNetV3 as the base network. The code uses the OpenCV library for image and video processing.The requirements.txt file contains a list of the Python libraries required to run the object detection code. Let’s install libraries using the following command.

pip install -r requirements.txt

The Single Shot Detector (SSD) Algorithm

The Single Shot Detector (SSD) algorithm is a popular deep learning-based approach for object detection. It is a one-stage detector that directly predicts the bounding boxes and class labels for all objects in an image, without using a separate region proposal network. This makes it faster and more efficient than other object detection algorithms.The SSD algorithm uses a convolutional neural network (CNN) to extract features from the input image, and then applies a set of detection heads to predict the class labels and bounding boxes for each object. The MobileNetV3 architecture is a popular choice for the base network in SSD, as it is lightweight and optimized for mobile devices.

Running Object Detection with OpenCV

To perform object detection with OpenCV, we first need to install the necessary libraries. We can do this by running the following commands in the terminal:

git clone https://github.com/brainypi/brainypi-opencv-examples.git
cd object-detection
pip install -r requirements.txt

Once the libraries are installed, we can run the object detection code by running the following command:

python objectDetection.py

This will start the object detection script, which captures frames from the camera and uses the SSD model to detect objects within the frames. The detected objects are then displayed on the screen in real-time, with bounding boxes around them.To exit the program, press the q key.
Adapting the Code for Your Own Use

If you want to use this face recognition code for your own project, you’ll need to make some changes to customize it for your specific use case. Here are some tips for adapting the code:

`detectObjects()` – Function which is responsible for detection objects

- Inputs:
  - img ([cv2.Mat]): Input Image
  - thres ([int]): Object detection threshold
  - draw (bool, optional): Draw binding boxes in output frame. Defaults to True.
  - objects (list, optional): List of objects to filter, i.e will only detect these objects. Defaults to [].

- Outputs:
  - img [cv2.Mat]: Output image
  - objectInfo [list]: Object information

We can use this function as follow

threshold = 0.45 
result, objectInfo = detectObjects(img, threshold, 0.2)

Let’s change the object detection threshold to remove false positives.

threshold = 0.85 
result, objectInfo = detectObjects(img, threshold, 0.2)

Now it will only detect objects which the model is 85% sure of the object.
Let’s say you want to detect only 1 or 2 objects like chair or books, then you can

threshold = 0.85 
objects = ["chair", "books"] 
result, objectInfo = detectObjects(img, threshold, 0.2, true, objects)

so will only detect chair and books and filter out all the other objects.Following example is for Cell phone and Person detection.

You can see the full example in the code here — https://github.com/brainypi/brainypi-opencv-examples/blob/main/object-detection/objectDetection.py

Conclusion

Object detection using OpenCV on brainypi can be used for many Edge applications. In this tutorial, we explored how to perform real-time object detection using Python, OpenCV, and the SSD algorithm. We also discussed some improvements to the code. With this knowledge, you can develop your own object detection system on Brainy Pi. Checkout more opencv applications