HuskyLens – An AI Camera: Click, Learn, and Play | Vision Sensor | DFRobot Electronics
Curious what experiences are out there.
I have not tested it for an FRC robot application but for small scale robotics and/or a classroom educational setting, I think it’s fantastic, especially for what it costs. I was impressed with what it could do. Very easy to set up and use, especially for someone who has never done anything with vision before. You don’t have quite the range of custom setting options like you would on other options like a Limelight, but I’ve found most of the built in modes on the HuskyLens to work pretty well, at least at close range. My tinkering was more with the device itself, and not as much with integration with a robot, but the Arduino support and documentation seems decent. You should be able to get your relevant data out of it to any controller over a serial line. I had planned on using the HuskyLens with Pololu Zumo robots, but ran short on time. The DFRobot Maqueen line of robots is a good companion for the HuskyLens too. I like that the HuskyLens is supported in Microsoft MakeCode.
I’d say it’s worth the $50 to pick one up and try it. But, I’m not sure if it’s the right thing for many FRC applications where you’re trying to target at substantial range and you often have varying lighting conditions. But for FTC, I could see this thing being quite useful.
Reminds me of the JeVois when it first came out. I’m considering getting one, but the only interface being UART/I2C makes it annoying to interface with the RoboRIO. It probably wouldn’t be very effective on its own.
That said, it does look like an incredibly fun learning tool.
Also a note, Digikey and Mouser stock them, and I would recommend the kit with the silicone case.
I just noticed the Husky Lens. I thought I would revive this topic to see if anyone else has tried it with FRC or just played with it.
I tried the HuskyLens with AprilTags. Tracking the Tag36h11 family works pretty great for this little USB-powered $50 device. Unfortunately, the Tag16h5 family that FIRST switched to for 2023 appears to not be supported at all by the HuskyLens. You can teach it as an object rather than as a tag, and it sort of works, but you’re limited to one tag/object, and it will not distinguish between different tags.
If FIRST had stuck with the Tag36h11 family, this could have been a potentially viable ultra-low cost vision solution that could potentially even see use in FTC. But, other solutions out there, are much better performing anyhow, and I guess it’s not a bad thing to put the emphasis on performance.
Why did this need to be revived right before I spend money on robot parts…
I really wish the firmware for this thing was open source, but no such luck. I like the hardware and I feel it could do more than it does in the current firmware.
This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.