This Bio-Inspired 3D Camera Cracks the Sub-Watt Barrier

Mapping out a 3D scene requires a lot of equipment and computations. Traditional solutions commonly rely on multiple cameras, specialized sensors, or powerful GPUs crunching through mountains of data just to figure out where objects are in space. They work quite well under most conditions, but they are bulky and draw a lot of power. These factors make them unsuitable for mobile applications in wearables or drones, for instance.

A group led by researchers at Northwestern University has taken a completely different approach to 3D mapping that is both tiny and energy efficient. Inspired by jumping spiders, the team has created a system called SpiderCam that borrows tricks from the animal kingdom to make mapping much easier. This technology could ultimately be deployed to miniature, battery-powered devices of all sorts.

Jumping spiders have an unusual visual system that allows them to estimate distances despite having brains no larger than a poppy seed. Instead of relying on stereo vision alone, their eyes contain multiple retinal layers that simultaneously capture images with different levels of focus. By comparing these subtle differences in blur, the spiders can determine how far away objects are before making jumps. SpiderCam applies the same concept in hardware.

The researchers’ custom camera captures two images of the same scene at slightly different focus settings using a beam splitter and a pair of low-power image sensors. A specialized depth-from-differential-defocus (DfDD) algorithm then analyzes the differences between the images and converts those blur variations into depth measurements. Unlike many modern computer vision systems, SpiderCam does not depend on complex neural networks or expensive active sensing hardware.

To keep power consumption exceptionally low, the researchers implemented the entire processing pipeline directly on a low-power FPGA rather than a conventional processor. The team also redesigned portions of the DfDD algorithm to reduce computational overhead, minimize memory usage, and improve robustness when working with noisy image sensors.

The prototype produces sparse depth maps with a resolution of 480 × 400 pixels at 32.5 frames per second while operating over a working range of approximately 52 centimeters. The complete system — including both image sensors and processing hardware — consumes just 624 milliwatts of power. According to the team, this makes SpiderCam the first passive FPGA-based 3D camera system to operate below one watt.

Because SpiderCam relies on passive imaging rather than projected light, it can handle reflective surfaces, transparent objects, and moving scenes that often challenge active depth-sensing systems. The approach also eliminates the need for power-hungry illumination hardware, making it particularly attractive for mobile platforms.

The researchers envision future versions of SpiderCam finding their way into wearable devices, assistive technologies, small robots, drones, and augmented reality systems. They also plan to improve the optics, widen the field of view, and eventually develop custom chips that could reduce power consumption even further.