OpenSource GUI Tool For OpenCV And DeepLearning

February 28, 2020 by Inderpreet Singh 4 Comments

AI and Deep Learning for computer vision projects has come to the masses. This can be attributed partly to the community projects that help ease the pain for newbies. [Abhishek] contributes one such project called Monk AI which comes with a GUI for transfer learning.

Monk AI is essentially a wrapper for Computer Vision and deep learning experiments. It facilitates users to finetune deep neural networks using transfer learning and is written in Python. Out of the box, it supports Keras and Pytorch and it comes with a few lines of code; you can get started with your very first AI experiment.

[Abhishek] also has an Object Detection wrapper(GitHub) that has some useful examples as well as a Monk GUI(GitHub) tool that looks similar to the tools available in commercial packages for running, training and inference experiments.

The documentation is a work in progress though it seems like an excellent concept to build on. We need more tools like these to help more people getting started with Deep Learning. Hardware such as the Nvidia Jetson Nano and Google Coral are affordable and facilitate the learning and experimentation.

Relive The Glory Days Of Cable TV With This Retro Weather Feed

February 21, 2020 by Dan Maloney 43 Comments

This may surprise younger readers, but there was once a time when the reality programming on The Weather Channel was simply, you know, weather. It used to be no more than a ten-minute wait to “Local on the Eights”, with simple text crawls of local conditions and forecasts that looked like they were taken straight from the National Weather Service feed. Those were the days, and sadly they seem to be gone forever.

Or perhaps not, if this retro weather channel feed has anything to say about it. It’s the product of [probnot] and consists of a simple Python program that runs on a Raspberry Pi. Being from Winnipeg, [probnot] is tapping into Environment Canada for local weather data, but it should be easy enough to modify to use your local weather provider’s API. The screen is full of retro goodness, from the simple color scheme to the blocky white text; the digital clock and local news crawl at the bottom complete the old school experience. It doesn’t appear that the code supports the period-correct smooth jazz saxophone, but that too should be a simple modification.

All jibing aside, this would be a welcome addition to the morning routine. And for the full retro ride, why not consider putting it in an old TV case?

Continue reading “Relive The Glory Days Of Cable TV With This Retro Weather Feed” →

An Eight-Day Home Automation Hackathon Is Inspiration For Getting More Projects Done

January 24, 2020 by Mike Szczys 11 Comments

There’s nothing quite like a deadline to cut through extras and get right at the heart of the problem. Maybe we should all follow Interpreet’s example and stop thinking about automating our homes and just make it in an eight-day hackathon. His talk at the 2019 Hackaday Superconference covers the zero-to-deployment home automation build he finished in the eight days leading up to his move from one continent to another.

Hackaday’s very own Inderpreet Singh found himself pulling up roots and moving from his home in India to teach at Centennial College in Toronto, Canada. He needed a way to keep an eye on his home from afar and the name of the game is IoT. When the only choice is “whatever works right now”, you can learn a lot about simple solutions.

He chose familiar hardware to work with, with the ESP8266 making up the bulk of the nodes and a Raspberry Pi as as a central hub for the setup. He chose to communicate between all the nodes on his system using WiFi because the hardware is robust and available. With security in mind, he keeps the automation system separate from the daily use WiFi system by grabbing an extra access point to serve as the automation network. The Raspberry Pi serves as a router of sorts; its Ethernet port is connected to the IoT device’s AP, while the onboard WiFi is used to connect to the home’s main AP for a connection to the wider Internet.

Software for the system is built on a REST API served by a Python Flask app. Many would advocate for using MQTT but Inderpreet’s testing with that protocol came up short as the broker he intended to use was no longer available. One of the interesting parts of his system design is that all nodes will check in at regular intervals; this allows them to inquire about actions they need to take, but it also allows the system to detect a malfunctioning node immediately. I’ve seen a similar trick used by Elliot Williams where he assigns a “ping” topic to all MQTT devices that causes them to report in with their IP address. Having a system to query and ensure the health of every node is a big tip to take away from this talk.

Continue reading “An Eight-Day Home Automation Hackathon Is Inspiration For Getting More Projects Done” →

Hackaday Links: January 19, 2020

January 19, 2020 by Dan Maloney 20 Comments

We’ve seen some interesting pitches in personal ads before, but this one takes the cake. Japanese billionaire Yusaku Maezawa is looking for a date to go along with him on his paid trip to the Moon, with the hope of finding a life partner. Maezawa is slated to be SpaceX’s first commercial lunar flyby customer, and will make the trip no earlier than 2023. That should give him plenty of time to go through the 20,000 applications he received from single women 20 and older with bright personalities and positive attitudes. And he should have plenty of time to make an awesome mixtape for the ride.

Imagine snooping through your kid’s garbage can only to find a used syringe lying in there. Most of us would likely be able to tell that the syringe once contained thermal compound or solder paste and be suitably proud of the little chip off the block, but apparently Cooler Master has fielded enough calls from panicked normie parents that they decided to change the design of their applicators. Given the design of the new applicator we doubt that’s really the reason, but it’s a good marketing story, and we can totally see how someone could mistake the old applicator for something illicit.

It looks as though SpaceX could be getting itself into legal trouble with its Starlink launches. Or more correctly, the FCC might, having apparently violated the National Environmental Policy Act, a Nixon-era law that requires government agencies to consider the environmental impact of any projects they approve. The Federal Communications Commission has been using a loophole in the law to claim a “categorical exemption” from these reviews when approving communications projects, particularly space-based projects. It’s not clear whether space is legally considered part of the environment, so the lawyers are hashing that out. If the FCC gets sued and loses, it’s not clear what happens to the existing Starlink satellites or future launches. Stay tuned for details.

Don’t forget that the Open Hardware Summit is coming soon. The 2020 meeting is the 10th anniversary of the confab, to be held on March 13 in New York. Hackaday is, of course, a proud sponsor of the conference, and our own Sophi Kravtiz will be the keynote speaker! Get your tickets soon.

Tired of off-loading data manipulation and analysis tasks to R in your Python programs? Then you’re probably already aware of Pandas, the Python library that converts data into dataframe objects for easier manipulation. Pandas has (have?) been in pre-release for years, but there’s now a legit 1.0.0 release candidate available. Now might be the time for you Python data mungers to get onboard the Pandas Express.

And finally, the Consumer Electronics Show is a yearly gift to anyone in the tech media, providing as it does so many examples of outrageous uses for the latest technology. To wit, we have LuluPet, the world’s first feces-analyzing cat litter box. LuluPet uses a built-in camera along with IR sensors and an “AI chip” to monitor your cat’s dookie and provide an alert if anything looks awry. On the one hand, inspecting cat poop is a job we’d love to outsource, but on the other hand, most cats we know are quick to cover the evidence of their excretions with kitty litter, leaving a clay-encrusted blob rather than the turds with defined borders that would seem to be needed for image recognition to do its job. We’ll reserve judgment on this one until we see a review.

Bask In The Glory Of This 336 LED Digit Display

January 18, 2020 by Tom Nardi 12 Comments

[Chris Combs] recently took the wraps off of an incredible art piece that he calls Road Ahead which uses 336 seven segment LED digits to create an absolutely gorgeous display. With a piece of smoked acrylic to slightly diffuse the orange glow of the LEDs, the end result has a distinctively retro look that we’d gladly spend all day staring at.

For those looking to dig a bit deeper, [Chris] has put together some very impressive documentation over on Hackaday.io that goes into plenty of detail on how he designed and built this beauty. From the design of the PCBs that carry all of the 0.3″ SMD displays to the custom software running on the Raspberry Pi 3 that powers it, there’s no technical stone left unturned.

According to the build log, this is the second version of the display. The first one was housed in a rather attractive wooden enclosure, but as [Chris] explains, that was precisely the problem. He wanted something that looked cold and unfeeling as the nearly 340 digits flashed away with potentially ominous intent. So he ditched the wooden case for a powder coated steel one that looks more like the front panel of a mainframe than something you’d pick up at the craft store.

Another interesting point explained in the write-up is how the Python software is designed to treat the hardware as a contiguous graphical display rather than just an array of independent digits. Grayscale images can be reproduced on the by using PWM to adjust the brightness of each segment’s corresponding “pixel”; though admittedly it takes a bit of imagination to see the intended image with a resolution this low.

This project reminds us of the incredible LED hexdump display we saw not that long ago, down to the PWM trickery for squeezing “graphics” out of these exceptionally non-graphical elements. With any luck, perhaps these are the opening shots in an arms race to see who can build the largest array of multi-segment LED displays.

New Part Day: Arduino Goes Pro With The Portenta H7

January 15, 2020 by Tom Nardi 92 Comments

The Consumer Electronics Show in Las Vegas is traditionally where the big names in tech show off their upcoming products, and the 2020 show was no different. There were new smartphones, TVs, and home automation devices from all the usual suspects. Even a few electric vehicles snuck in there. But mixed in among flashy presentations from the electronics giants was a considerably more restrained announcement from a company near and dear to the readers of Hackaday: Arduino is going pro.

While Arduino has been focused on the DIY and educational market since their inception, the newly unveiled Portenta H7 is designed for professional users who want to rapidly develop robust hardware suitable for industrial applications. With built-in wireless hardware and the ability to run Python and JavaScript out of the box, the powerful dual-core board comes with a similarly professional price tag; currently for preorder at $99 USD a pop, the Portenta is priced well outside of the company’s traditional DIY and educational markets. With increased competition from other low-cost microcontrollers, it seems that Arduino is looking to expand out of its comfort zone and find new revenue streams.

Continue reading “New Part Day: Arduino Goes Pro With The Portenta H7” →

This Week In Security: Camera Feeds, Python 2, FPGAs

January 10, 2020 by Jonathan Bennett 6 Comments

Networked cameras keep making the news, and not in the best of ways. First it was compromised Ring accounts used for creepy pranks, and now it’s Xiaomi’s stale cache sending camera images to strangers! It’s not hard to imagine how such a flaw could happen: Xiaomi does some video feed transcoding in order to integrate with Google’s Hub service. When a transcoding slot is re-purposed from one camera to another, the old data stays in the buffer until it is replaced by the new camera’s feed. The root cause is probably the same as the random images shown when starting some 3D games.

Python is Dead, Long Live Python

Python 2 has finally reached End of Life. While there are many repercussions to this change, the security considerations are important too. The Python 2 environment will no longer receive updates, even if a severe security vulnerability is found. How often is a security vulnerability found in a language? Perhaps not very often, but the impact can be far-reaching. Let’s take, for instance, this 2016 bug in zipimport. It failed to sanitize the header of a ZIP file being processed, causing all the problems one would expect.

It is quite possible that because of the continued popularity and usage of Python2, a third party will step in and take over maintenance of the language, essentially forking Python. Unless such an event happens, it’s definitely time to migrate away from Python2.
Continue reading “This Week In Security: Camera Feeds, Python 2, FPGAs” →