Is it really cheating if the aimbot you’ve built plays the game worse than you do?
We vote no, and while we take a dim view on cheating in general, there are still some interesting hacks in this AI-powered bot for Valorant. This is a first-person shooter, team-based game that has a lot of action and a Counter-Strike vibe. As [River] points out, most cheat-bots have direct access to the memory of the computer which is playing the game, which gives it an unfair advantage over human players, who have to visually process the game field and make their moves in meatspace. To make the Valorant-bot more of a challenge, he decided to feed video of the game from one computer to another over an HDMI-to-USB capture device.
The second machine has a YOLOv5 model which was trained against two hours of gameplay, enough to identify friend from foe — most of the time. Navigation around the map was done by analyzing the game’s on-screen minimap with OpenCV and doing some rudimentary path-finding. Actually controlling the player on the game machine was particularly hacky; rather than rely on an API to send keyboard sequences, [River] used a wireless mouse dongle on the game machine and a USB transmitter on the second machine.
The results are — iffy, to say the least. The system tends to get the player stuck in corners, and doesn’t recognize enemies that pop up at close range. The former is a function of the low-res minimap, while the latter has to do with the training data set — most human players engage enemies at distance, so there’s a dearth of “bad breath range” encounters to train to. Still, we’re impressed that it’s possible to train a machine to play a complex FPS game at all, let alone this well.
Were you aware that there’s a market for backpack-housed live streaming video systems, and that they can cost as much as $1600? Apparently these things are popular with social media moguls who want to stream themselves living their fabulous lives to people sitting at home watching on YouTube or Twitch. But believing that even slack jawed yokels like us should have access to the same technology, [Speedify Labs] has been working on less expensive DIY alternative based on the Raspberry Pi 4.
Now you’ll note we didn’t use the term “cheap” to describe this build. As detailed here, it’s still going to cost you around $600. You could always swap out the Sony AS-300 camera and Elgato Cam Link capture device with cheaper versions, but the goal of this project was to deliver high quality HD video that’s comparable to what the professional rigs are capable of, so those kinds of concessions were avoided.
Whatever video source your audience and budget are comfortable with, it eventually gets fed into the Raspberry Pi 4 which uses an
ffmpeg one-liner to encode the video and ultimately push it out as 720p at 24 FPS, which [Speedify Labs] says seems to be about as good as the Pi can do. The operator is able to start and stop the stream at will using a Circuit Playground Express and a Python script.
Of course, the trick to all of this is getting the video stream uploaded over potentially flaky mobile networks. But as you might have guessed, that’s where [Speedify Labs] gets to flex their eponymous product: a VPN with software channel bonding that allows you to combine multiple Internet connections for higher bandwidth and reliability. With their software, the Pi is able to stream the video through two mobile phones connected to it over USB. As demonstrated in the video below, the setup was able to maintain the stream even as they walked in and out of buildings.
Our very own [Lewin Day] wrote about his experiments with streaming video over 4G on the Raspberry Pi which might be of interest to anyone looking to take their show on the road. Though if you want to get serious it would be worth taking a look at the impressive mobile streaming rig that [Jenny List] saw at the BornHack 2019 hacker camp in Denmark.
Continue reading “A Raspberry Pi 4 Video Streaming Backpack”
Autodetection of hardware was a major part of making computers more usable for the average user. The Amiga had AutoConfig on its Zorro bus, Microsoft developed Plug And Play, and Apple used NuBus, developed by MIT. It’s something we’ve come to take for granted in the modern age, but it doesn’t always work correctly. [Evan] ran into just this problem with a video capture card that wouldn’t autodetect properly under Linux.
The video capture card consisted of four PCI capture cards with four inputs each, wired through a PCI to PCI-E bus chip for a total of sixteen inputs. Finding the cause of the problem wasn’t too difficult – the driver was detecting the card as a different model with eight inputs, instead of the sixteen inputs actually present on the card. The driver detects the device plugged in by a unique identifier reported by the card. The code on the card was identical to the code for a different model of card with different hardware, causing the issue.
As a quick test, [Evan] tried fudging the driver selection, forcing the use of a driver for a sixteen-input model. This was successful – all sixteen inputs could now be used. But it wasn’t a portable solution, and [Evan] would have to remember this hack every time the card needed to be reinstalled or moved to a different computer.
Looking further at the hardware, [Evan] discovered the card had four 24c02 EEPROM chips on board – one for each PCI card on board. Dumping the contents, they recognised the unique identifier the driver was using to determine the card’s model. It was then a simple job to change this value to one that corresponded with a sixteen-input card to enable functional autodetection by burning a new value to the EEPROM. [Evan] then published the findings to the LinuxTVWiki page. Continue reading “EEPROM Hack To Fix Autodetection Issues”
There’s a number of devices out there that extend HDMI over IP. You connect a video source to the transmitter, a display to the receiver, and link the two with a CAT5/5e/6 cable. These cables are much cheaper than HDMI cables, and can run longer distances.
[Daniel] didn’t care about extending HDMI, instead he wanted a low cost HDMI input for his PC. Capture cards are a bit expensive, so he decided to reverse engineer an IP HDMI extender.
After connecting a DVD player and TV, he fired up Wireshark and started sniffing the packets. The device was using IP multicast on two ports. One of these ports had a high bitrate, and contained JPEG headers. It looked like the video stream was raw MJPEG data.
The next step was to write a listener that could sniff the packets and spit the data into a JPEG file. After dealing with some quirks, JPEG images could be saved from the remote device. Some more code was needed to have the computer initiate the streaming, and to extract audio from the second port.
In the end, video capture with the low cost device was possible. [Daniel] also provided a bonus teardown of the device in his writeup.
[Carlos Agell] sent in a tip where he captured images from an analog camera with an Arduino.
We’ve seen a few AVR/Arduino hacks that generate video, although overclocking is necessary if you want to do anything beyond a Breakout clone. [Carlos]’ hack bucks that trend and now he can capture video with an Arduino.
The project captures individual frames from NTSC video at a resolution of 128×96. Although the Arduino isn’t powerful enough for real-time capture, [Carlos] managed this by capturing only thresholds and sending them over to a computer running a program coded in LabVIEW. The PC program reassembles the images of the thresholds and produces a tiny image in 3-bit grayscale.
[Carlos] used the Video Experimenter shield which is impressive in it’s own right. The Video Experimenter is able to do object tracking and edge detection, so we’re wondering when we’ll see robots with computer vision running off an Arduino. Check out a demo of the nootropic design video experimenter shield after the break.
UPDATE: Carlos wrote a sketch in Processing that does the same thing as his LabVIEW program.
Continue reading “Capturing Video With An Arduino”