Steven Cheng, who works in computer vision and robotics, built a system that uses deep learning together with a laser-based targeting device to hunt mosquitoes. The setup forms a closed-loop: detection, identification and targeting occur in real time between camera, neural network and actuation hardware.
Hardware and training
The primary sensor was a DSLR camera fitted with a high-magnification zoom lens, which Cheng used both to collect training data and as an operational sensor. He captured detailed footage of mosquitoes and trained a custom neural network on that dataset so the model could recognize flying insects.
The same camera monitors the space during operation; the trained deep-learning model detects mosquitoes and a precision pan/tilt mechanism with an attached laser locks on and destroys the insects. According to TechSpot, this active tracking and targeting contrasts with conventional passive, attraction-based traps.
Safety measures
Cheng also implemented safety features: a second wide-angle camera scans for people and flammable objects. If such targets enter the firing line, the system disables the shot to reduce the risk of accidental harm.
Results and public reaction
After assembling and testing the system, Cheng deployed it in his apartment and claims that it cleared all mosquitoes from his home in a single night. He shared progress updates on social media, including a tweet dated May 28, 2026, noting he spent about four months building the device.
Reactions online were mixed: some praised the ingenuity of the engineering, while others warned that an indoor, automatically targeting laser still poses risks of misidentification or eye injury despite built-in safety measures.
Why it matters
The project illustrates how machine vision and robotics can be combined to address practical problems, but it also highlights safety and ethical concerns when potentially hazardous devices are used in domestic settings.
