Tag Archives: camera board

Now with added cucumbers

via Raspberry Pi

Working here at Pi Towers, I’m always a little frustrated by not being able to share the huge number of commercial businesses’ embedded projects that use Raspberry Pis. (About a third of the Pis we sell go to businesses.) We don’t get to feature many of them on the blog; many organisations don’t want their work replicated by competitors, or aren’t prepared for customers and competitors to see how inexpensively they’re able to automate tasks. Every now and then, though, a company is happy to share what they’re using Pis for.


Makoto Koike, centre, with his parents on the family cucumber farm

Here’s a great example: a cucumber farm in Japan, which is using a Raspberry Pi to sort thorny cucumbers, saving the farmer eight to nine hours’ manual work a day.

Makoto Koike is the son of farmers, who works as an embedded systems designer for the Japanese car industry. He started helping out at his parents’ cucumber farm (which he will be taking over when they retire), and spotted a process that was ripe for automation.


Cucumbers from the Makotos’ farm

At the Makotos’ farm, cucumbers are graded into nine categories: the straightest, thickest, freshest, most vivid cucumbers (which must have plenty of characteristic spurs) are the best, and can be sold at the highest price. Makoto-san’s mother was in charge of sorting the cucumbers every day, which took eight hours at the peak of the harvest. Makoto-san had an epiphany after reading about Google’s AlphaGo beating the world number one professional Go player. He realised that machine learning and deep learning meant the sorting process could be automated, so he built a process using Google’s open-source machine learning library, TensorFlow, and some machinery to process the cucumbers into graded batches.


Sorting in action


Camera interface

Google have put together a diagram showing how the system works:


There are difficulties in building this sort of system, not least the 7000 cucumbers, pre-graded by his mother, that Makoto-san had to photograph and label over a period of three months to give the model material to train with. He says:

“When I did a validation with the test images, the recognition accuracy exceeded 95%. But if you apply the system with real use cases, the accuracy drops down to about 70%. I suspect the neural network model has the issue of “overfitting” (the phenomenon in neural networks where the model is trained to fit only the small training dataset) because of the insufficient number of training images.”

Still, it’s an impressive feat, and a real-world >95% accuracy rate is not unfeasible with a big enough data set. We’d be interested to see how the setup progresses, especially as more automation is added; right now, cucumbers are added to the processing hopper by hand, and a human has to interact with the touchscreen grading panel. Here’s the system at work:

TensorFlow powered cucumber sorter by Makoto Koike

Uploaded by Kazunori Sato on 2016-08-05.

We’re hoping to see some updates from the Makoto family as the system evolves. And in the meantime, if you have an embedded project you’d like to share with us, let us know in the comments!


The post Now with added cucumbers appeared first on Raspberry Pi.

Identifying The Hallway Whistler

via Raspberry Pi

Becky Stern suffers from that same condition that many of us apartment dwellers are affected by: a curiosity about who is making noise outside the door.

Living within a large New York City apartment, Becky wanted to be able to see out of her peep hole without having to leave her desk. After all, the constant comings and goings of any shared property, though expected, can often be distracting.

(And seriously, whoever keeps slamming their door in my apartment block at 4am WILL suffer my wrath!)

So she decided to use a motion detector to trigger a Pi camera at her door. The camera would then stream live video back to a monitor within her apartment: a wireless peep hole, allowing her the freedom to be productive without having her eye to the door.

Peep Hole Cam

Becky used a Pi Zero for the project and took to the internet to educate herself on how to code a live streaming camera with motion detection. Tony D’s Cloud Cam tutorial gave her everything she needed to get the project working… and a handful of magnets, plus an old makeup bag, finished of the job.

Pi Zero Peep Hole Camera

Tutorial: http://www.instructables.com/id/Pi-Zero-Peep-Hole-Camera/ Subscribe for new videos Mondays and Thursdays! http://www.youtube.com/user/bekathwia previous video: https://youtu.be/p7uUcNFfP3Q tech playlist: https://www.youtube.com/playlist?list=PLxW5bBHPfdBzmynozxfEPv2DJgyoFiqgn this time last year: https://youtu.be/kZmyXzzXqfc Connect with Becky: http://www.instructables.com/member/bekathwia https://twitter.com/bekathwia http://instagram.com/bekathwia http://bekathwia.tumblr.com/ http://www.pinterest.com/bekathwia/ https://www.snapchat.com/add/bekathwia tip jar: https://www.patreon.com/beckystern Music is “Marxist Arrow” from the YouTube Music Library

Along with live streaming, the camera could be set up to take and upload photos and video to a cloud server; a handy tool to aid in home security. Taking the project further afield, she could allow remote access to the camera, allowing her to view the hallway while away from home. Did the delivery man leave your expected package? Which of the neighbours kids is the one trailing mud across the carpet?

And seriously… who keeps whistling every time they come home?!

The post Identifying The Hallway Whistler appeared first on Raspberry Pi.

Camera board comparisons: Pi NoIR v1 vs Pi NoIR v2

via Raspberry Pi

The new 8 megapixel Raspberry Pi camera board has been out for just over a month, and we’ve been seeing some really impressive work being done with it. As many of you know, we also make a version of the camera board with no infrared (IR) filter: the Pi NoIR. Version 2 of the Pi NoIR has also been upgraded to use an 8 megapixel Sony sensor. People use the Pi NoIR to see in the dark (especially useful for wildlife camera traps and for security cams), to achieve some wacky camera effects, and to work on hyperspectral imaging.

What we haven’t seen so far is any comparison of the output from the Pi NoIR v1 and the Pi NoIR v2. So we were really pleased to find this video from Andr.oid Eric which demonstrates the cameras’ raw output side by side, alongside output from both cameras with a selection of IR and UV filters.

Compare Pi Camera NoIR v1 vs V2

Compare Pi Camera NoIR v1 vs V2 http://helloraspberrypi.blogspot.com/2016/05/compare-pi-camera-noir-v1-vs-v2.html

Let us know in the comments if you have your own comparison photos or video from v1 and v2 of either of the camera boards – we’d love to see more!

The post Camera board comparisons: Pi NoIR v1 vs Pi NoIR v2 appeared first on Raspberry Pi.

Le Myope – a confused camera

via Raspberry Pi

This is very silly indeed.

Salade Tomate Oignon in Paris seems to be making a bit of a habit of doing outlandish things with Raspberry Pi and other people’s photography. You might remember Layer Cam from a couple of years ago, which allows you to point a sandwich box pretending to be a camera at a landmark and serves up somebody else’s picture of the same thing, using GPS coordinates and Google Image Search.

His newest Raspberry Pi hack, Le Myope (for non-Francophones, that’s The Shortsighted), actually includes a camera – but the results are not what you’d expect. Here’s a bit of video to show you more.

Le myope: a similar images Raspberry Pi camera

Short-sighted camera based on a Raspberry Pi and Google similar images. Find instruction and code to build your own: http://saladetomateoignon.com/Wordpress/a-short-sighted-raspberry-pi-camera/ Music: Samuel Belay – Qeresh Endewaza Logo: Alice www.alicesawicki.com Images: Charly www.nnprod.com

Salade Tomate Oignon says:

Even more imprecise than a blurry polaroid picture, or than a filter-abused instagram shot.
Using the most advanced algorithms based on machine learning and computer vision, here is ‘Le myope’, a short-sighted camera.
The new iteration of the layercam ‘Why are you taking this picture? It’s already on the Internet!’ is a Raspberry Pi based camera, that takes a picture and returns a similar one from Google similar image search.
Use it in a popular place and chances are that you will get the same picture taken by someone else. (That happened with the mural during one of the tests)
Use it in a remote place and get random roughly similar pictures from all over the internet!

This is an extremely daft project which pleased us out of all proportion. You can find code and instructions to build your own at Salade Tomate Oignon’s website. Go forth and take other people’s photographs.


The post Le Myope – a confused camera appeared first on Raspberry Pi.

Learn all about the new Raspberry Pi Camera Module v2 in The MagPi 45

via Raspberry Pi

Earlier this week, the brand new Raspberry Pi Camera Module v2 was revealed to the world, its headline feature being an 8-megapixel sensor. It’s been a few years since the original came out and the new camera is an excellent little upgrade to the existing model; you can find out all the details in our complete breakdown in Issue 45 of The MagPi magazine, which is out today.

Picture perfect, the new Pi Camera Module v2

Picture perfect, the new Pi Camera Module v2

As well as covering the camera and giving you some projects to start you off with it, we also have a look at the ten best Pi-powered arcade machines, which should give you some ideas for a retro games system of your own. There are also tutorials on creating lighting effects for costumes with a Pi and some NeoPixels, making an Asteroids clone in Basic, and building an IoT thermometer. We also have Astro Pi news, excellent projects, reviews, and everything else you’d expect from your monthly MagPi.

A model railway, in-part powered by Pi Zero

A model railway, powered in-part by Pi Zero

Highlights from issue 45:

  • Replicate an Astro Pi experiment
    Create a humidity sensor, similar to the Sweaty Astronaut code
  • Hacking with dinosaurs
    The MagPi heads to the Isle of Wight to see how some animatronic dinos are being hacked with Pi
  • Original games on the Pi
    Play three brand-new games on your Pi thanks to YoYo and GameMaker Studio
  • Moon pictures
    Find out how to use the camera board to take amazing photos of the moon
  • And much, much more!

How to buy
As usual, you can get The MagPi in store from WH Smith, Tesco, Sainsbury’s, and Asda as well as buying copies online from our store. It’s also available digitally via our app on Android and iOS. If you fancy subscribing to the magazine to make sure you never miss an issue, you can do that to on our subscription site.

Free Creative Commons download
As always, you can download your copy of The MagPi completely free. Grab it straight from the issue page for The MagPi 45.

Don’t forget, though, that like sales of the Raspberry Pi itself, all proceeds from the print and digital editions of the magazine go to help the Foundation achieve its charitable goals. Help us democratise computing!

We hope you enjoy this month’s issue! Before anyone asks, no, the magazine unfortunately does not come with a free camera. Sorry!

The post Learn all about the new Raspberry Pi Camera Module v2 in The MagPi 45 appeared first on Raspberry Pi.

New 8-megapixel camera board on sale at $25

via Raspberry Pi

The 5-megapixel visible-light camera board was our first official accessory back in 2013, and it remains one of your favourite add-ons. They’ve found their way into a bunch of fun projects, including telescopes, kites, science lessons and of course the Naturebytes camera trap. It was soon joined by the Pi NoIR infrared-sensitive version, which not only let you see in the dark, but also opened the door to hyperspectral imaging hacks.

As many of you know, the OmniVision OV5647 sensor used in both boards was end-of-lifed at the end of 2014. Our partners both bought up large stockpiles, but these are now almost completely depleted, so we needed to do something new. Fortunately, we’d already struck up conversation with Sony’s image sensor division, and so in the nick of time we’re able to announce the immediate availability of both visible-light and infrared cameras based on the Sony IMX219 8-megapixel sensor, at the same low price of $25. They’re available today from our partners RS Components and element14, and should make their way to your favourite reseller soon.

Visible light camera v2

The visible light camera…

...and its infrared cousin

…and its infrared cousin

In our testing, IMX219 has proven to be a fantastic choice. You can read all the gory details about IMX219 and the Exmor R back-illuminated sensor architecture on Sony’s website, but suffice to say this is more than just a resolution upgrade: it’s a leap forward in image quality, colour fidelity and low-light performance.

VideoCore IV includes a sophisticated image sensor pipeline (ISP). This converts “raw” Bayer-format RGB input images from the sensor into YUV-format output images, while correcting for sensor and module artefacts such as thermal and shot noise, defective pixels, lens shading and image distortion. Tuning the ISP to work with a particular sensor is a time-consuming, specialist activity: there are only a handful of people with the necessary skills, and we’re very lucky that Naush Patuck, formerly of Broadcom’s imaging team, volunteered to take this on for IMX219.

Naush says:

Regarding the tuning process, I guess you could say the bulk of the effort went into the lens shading and AWB tuning. Apart from the fixed shading correction, our auto lens shading algorithm takes care of module to module manufacturing variations. AWB is tricky because we must ensure correct results over a large section of the colour temperature curve; in the case of the IMX219, we used images illuminated by light sources from 1800K [very “cool” reddish light] all the way up to 16000K [very “hot” bluish light].

The goal of auto white balance (AWB) is to recover the “true” colours in a scene regardless of the colour temperature of the light illuminating it: filming a white object should result in white pixels in sunlight, or under LED, fluorescent or incandescent lights. You can see from these pairs of before and after images that Naush’s tune does a great job under very challenging conditions.

AWB with high colour temperature

AWB at higher colour temperature

AWB at lower colour temperature

AWB at lower colour temperature

As always, we’re indebted to a host of people for their help getting these products out of the door. Dave Stevenson and James Hughes (hope you and Elaine are having a great honeymoon, James!) wrote most of our camera platform code. Mike Stimson designed the board (his second Raspberry Pi product after Zero). Phil Holden, Shinichi Goseki, Qiang Li and many others at Sony went out of their way to help us get access to the information Naush needed to tune the ISP.

We’re really happy with the way the new camera board has turned out, and we can’t wait to see what you do with it. Head over to RS Components or element14 to pick one up today.

The post New 8-megapixel camera board on sale at $25 appeared first on Raspberry Pi.