A DIY laser turret for mosquitoes that begs you not to build it

What it does This hardware stack—Raspberry Pi or Jetson Nano, paired cameras, a DAC, and galvanometer mirrors—detects mosquitoes, hornets, or roaches in 3-D space, computes mirror angles with trigonometry, and aims a laser. The author repeatedly warns against using anything stronger than a 1 mW laser pointer because retinal damage is silent, irreversible, and reflections can blind bystanders outside the camera’s field of view. It is essentially a documented, academically backed bug-zapper control loop built from YOLOv4-tiny, OpenCV, and SPI commands to an mcp4922 DAC.

The interesting bit The README devotes more space to safety systems and published papers than to hype: the author has already sketched human-detection shutoffs, audio sensors, thermal cutoffs, and reflection detection, essentially treating the liability waiver as a core feature. That level of caution is the most unusual thing about the project.

Key highlights

• Stereo vision plus YOLOv4-tiny (or Haar cascades on slower boards) supplies X, Y, and Z coordinates for target triangulation.
• Benchmarked inference speeds on Jetson Nano: tkDNN hits 30–35 FPS, while plain Keras crawls at 4–5 FPS.
• Control path is fully documented: single-board computer → SPI → mcp4922 DAC → op-amp → galvanometer drivers at ±12 V.
• Includes links to two published papers: one on a low-cost stereovision disparity map, another on squeezing FPS out of embedded boards.
• Future roadmap covers lower-power eye-safe wavelength research, remote Wi-Fi control from a desktop GPU, and a pocket-sized custom PCB.

Caveats

• The author explicitly states the technology should not be used at home and cannot be made 100 % safe; reflected beams can hit people the cameras never see.
• Frame-rate constraints on single-board computers mean the system may struggle to track fast-moving targets in real time.
• Much of the repository appears to be documentation, schematics, and video demonstrations rather than a drop-in software package.

Verdict Worth studying if you are researching embedded computer vision, galvanometer control, or agricultural pest-management robotics. Skip it if you wanted a weekend backyard build—unless you enjoy reading safety disclaimers as a hobby.