How To Roll Your Own Custom Object Detection Neural Network

Real-time object detection, which uses neural networks and deep learning to rapidly identify and tag objects of interest in a video feed, is a handy feature with great hacker potential. Happily, it’s also possible to make customized CNNs (convolutional neural networks) tailored for one’s own needs, and that process just got easier thanks to some new documentation for the Vizy “AI camera” by Charmed Labs.

Raspberry Pi-based Vizy camera

Charmed Labs has been making hacker-friendly machine vision devices for a long time, and the Vizy camera impressed us mightily when we checked it out last year. Out of the box, Vizy has a perfectly functional object detector application that runs locally on the device, and can detect and tag many common everyday objects in real time. But what if that default application doesn’t quite meet one’s project needs? Good news, because it’s possible to create a custom-trained CNN, and that process got a lot more accessible thanks to step-by-step examples of training a model to recognize hands doing rock-paper-scissors.

Person and cat with machine-generated tags identifying them
Default object detection works well, but sometimes one needs custom results.

The basic process is this: Start with a variety of images that show the item of interest. Then identify and label the item of interest in each photo. These photos (a “training set”) are then sent to Google Colab, which will be used to generate a neural network. The resulting CNN model can then be downloaded and used, to see how well it performs.

Of course things rarely work perfectly the first time around, so at this point it’s pretty common for some refinement to be needed to increase accuracy. Luckily there are a number of tools to help do this without creating a new model from scratch, so it’s just a matter of tweaking until things perform acceptably.

Google Colab is free and the resulting CNNs are implemented in the TensorFlow Lite framework, meaning it’s possible to use them elsewhere. So if custom object detection has been holding up a project idea of yours, this might be what gets you over that hump.

Stadia Controller’s Two Extra Buttons Get Seen With WebHID

The Google Stadia game streaming service relied on a proprietary controller. It was a pretty neat piece of hardware that unfortunately looked destined for landfills when Google announced that Stadia would discontinue. Thankfully it’s possible to use them as normal gamepads, and related to that, [Thomas Steiner] has a developer blog post about how to talk to the Stadia controller via WebHID. Continue reading “Stadia Controller’s Two Extra Buttons Get Seen With WebHID”

Ban On Physical Mail Slated For NYC Jails, Which Could Go Digital Instead

Prison is a scary place, very much by design. It’s a place you end up when convicted of crimes by the judicial system, or in some cases, if you’re merely awaiting trial. Once you go in as a prisoner, general freedom and a laundry list of other rights are denied to you. New York City is the latest in a long list of municipalities looking to expand that list to include a ban on inmates receiving physical mail.

To achieve this, prisons across the US are instead switching to digital-only systems, which would be run by a private entity. Let’s look at the how, what, and why of this contentious new idea.

Continue reading “Ban On Physical Mail Slated For NYC Jails, Which Could Go Digital Instead”

A pair of PCBs with OLED character displays, showing a simple encryption program

The CryptMaster 2001 Provides Basic Lessons In Cryptography

Sending secret messages to your friends is fun, but today it’s so simple that you don’t even notice it anymore: practically any serious messaging system features encryption of some sort. To teach his kids about cryptography, [Michal Zalewski] therefore decided to bring the topic to life by building a handheld encryption system, called the CryptMaster 2001.

The system consists of an identical pair of hand-held devices built on prototype PCBs. A standard 16×2 character OLED display is used as an output device, which generates the ciphertext in real time as the plaintext is entered character by character through a rotary encoder. An ATmega328P manages the input and output routines and performs the encryption.

For ease of use, [Michal] wanted to use a reciprocal cipher, meaning one that uses the same operation for encryption and decryption. Trivial ciphers like ROT13 would be a bit too easy to crack, so he devised a slightly more complex system where each character in the input is encoded using a separate rearranged alphabet – a basic polyalphabetic substitution cipher.

[Michal]’s kids apparently had some good fun with the CryptMaster 2001, until his eldest son managed to reverse-engineer the encryption method, enabling him to decode messages without having access to one of the devices. This made the project a pretty decent lesson about the limits of basic cryptography: simply swapping letters doesn’t present a real challenge to anyone. Luckily, much more secure methods are available, even if you’re only using pen and paper.

Methane Pyrolysis: Producing Green Hydrogen Without Carbon Emissions

Generally, when we talk about the production of hydrogen, the discussion is about either electrolysis of water into oxygen and hydrogen, or steam methane reforming (SMR). Although electrolysis is often mentioned – as it can create hydrogen using nothing but water and electricity – SMR is by far the most common source of hydrogen. Much of this is due to the low cost and high efficiency of SMR, but a major disadvantage of SMR is that :slider

large amounts of carbon dioxide are released, which offsets some of the benefits of using hydrogen as a fuel in the first place.

Although capturing this CO2 can be considered as a potential solution here, methane pyrolysis is a newer method that promises to offer the same benefits as SMR while also producing hydrogen and carbon, rather than CO2. With the many uses for hydrogen in industrial applications and other fields, such as the manufacturing of fertilizer, a direct replacement for SMR that produces green hydrogen would seem almost too good to be true.

What precisely is this methane pyrolysis, and what can be expect from it the coming years?

Continue reading “Methane Pyrolysis: Producing Green Hydrogen Without Carbon Emissions”

A closeup of a black flexible PCB with an out-of-focus quarter in the background, approximately the same size as the end of the PCB we're looking at. One the right is a USB C connector and to its left are two SMD components with visible pins. Several smaller SMD components (resistors or caps?) are soldered to other parts of the board.

Making The AirPods Pro Case Repairable

Apple is often lauded for its design chops, but function is often sacrificed at the altar of form, particularly when repair is involved. [Ken Pillonel] has made it easier for everyone to replace the batteries or lightning port in the AirPods Pro case. (YouTube)

With such notable hacks as adding USB C to the iPhone already under his belt, [Pillonel] has turned his attention to fixing the notoriously poor repairability of AirPods and AirPods Pro, starting with the cases. While the batteries for these devices are available, replacement Lightning ports are not, and taking the housing apart for the case is an exercise in patience where the results can’t be guaranteed.

He designed a USB C replacement port for broken Lightning ports that is a perfect fit if you happen to get the case apart in one piece. If you’re less successful, he has you covered there too with a 3D printable enclosure replacement.

We sure miss the days of schematic proliferation here at Hackaday, but we know you don’t let glued enclosures or unobtainium parts stand in the way of repairs.

Continue reading “Making The AirPods Pro Case Repairable”

High-contrast pictures described on the article, put onto a wall beside a crib

High-Contrast Images For Hacker Family Harmonics

There’s a new addition to the Adafruit family, and it’s not a microcontroller board as you’d expect – however, we will still find plenty to learn from. On the Adafruit blog, [Phillip Torrone] shares a set of high-contrast images with us; the idea for such images is that they’re more appealing for a child during the first few months of its life, and not just that – they can support a kid’s development, too. The idea behind high-contrast images is twofold. During the first few months of life, a baby’s visual systems are only taking shape, and are nowhere near being advanced – so, sources of easily discernible and varied visual input can help it develop, as well as, perhaps, aid in holding attention.

The second part is – they look nice in their own way, and one would hope that a baby can appreciate them in the same way parents do. The images are quite varied, with some being somewhat electronics-themed (including an Adafruit logo, of course) and many being fairly neutral, which has to be an upside for us hackers when it comes to the spouse acceptance factor. For any of us interested, there are downloadable PDFs and

In a way, these are just like AprilTags – aiming to be helpful in development of visual algorithms. With such a family, we can’t wait to see what comes next – computer engineering books? Baby monitors with machine learning? Sleep-data-driven knit blankets? No matter what’s in store for us, we hope that for the Adafruit family, this journey will be smooth sailing.