Self-Powered Artificial Eye Mimics Human Vision for Ultra-Efficient Machine Learning and Motion Detection

Scientists Develop Self-Powered Artificial Eye Mimicking Human Vision

Researchers in Japan have created a self-powered artificial eye that replicates human vision with impressive precision. Using dye-sensitized solar cells, this device distinguishes colors with a resolution as fine as 10 nanometers in wavelength difference. It can perform logic operations and recognize motion and color with up to 82% accuracy by producing both positive and negative electrical signals depending on the light input.

Since it requires no external power and processes visual information similarly to a biological retina, this technology promises ultra-efficient machine vision systems suitable for autonomous vehicles, wearable health monitors, and remote sensors.

How the Artificial Eye Mimics Human Vision

This innovation, detailed in Scientific Reports, is based on an “optoelectronic artificial synapse”—a synthetic brain cell responsive to light. Much like how the human eye generates different electrical signals when seeing different colors, this artificial eye produces positive or negative electrical responses depending on the incoming light’s wavelength.

Current machine vision systems capture and store vast amounts of information at high frame rates (10 to 60 frames per second), demanding significant computational power. In contrast, human eyes selectively filter and compress visual data before sending it to the brain, which makes vision highly energy-efficient.

The Tokyo University of Science team addressed this by combining two dye-sensitized solar cells, each tuned to different parts of the spectrum: one sensitive to blue light and the other to red. When illuminated, the device outputs a positive voltage for blue wavelengths and a negative voltage for red. This bipolar response—being able to produce both positive and negative signals—is crucial, as traditional photodetectors only generate positive signals, limiting color discrimination abilities.

The device can distinguish wavelengths just 10 nanometers apart, rivaling human color discrimination. It also demonstrated the ability to perform basic logic functions essential for computing, handling these with ease due to its bipolar signal nature.

For a more advanced test, the researchers applied “physical reservoir computing,” where the device’s physical properties handle machine learning tasks instead of conventional processors. By flashing sequences of colored light representing different codes, the artificial eye differentiated patterns up to six digits long, recognizing 64 combinations—a notable achievement given the device’s simplicity.

In a practical demonstration, the device analyzed videos of six human motions—bending, waving one hand, waving both hands, jumping, running, and moving sideways—converted into red, green, and blue light pulses. It identified both motion and color with 82% accuracy, approaching the performance of more complex systems.

Why Is This Device So Efficient?

Because the device is based on solar cell technology, it produces its own electricity from the light it detects. This eliminates the need for external power sources and opens the door for autonomous sensors that can operate indefinitely under natural lighting.

Unlike conventional machine vision systems that require continuous power and extensive data processing, this bio-inspired design filters information directly in hardware. As a result, it needs less computing power, making it well suited for edge computing applications where devices operate independently from central servers.

Potential applications include compact surveillance cameras, medical devices monitoring vital signs with multi-wavelength light, environmental sensors in remote areas, and wearable health monitors. Although more complex visual scenes will require improved signal processing, optimizing dye materials and device architecture offers a path forward.

Summary of Research Methodology and Results

• Methodology: The device integrates two dye-sensitized solar cells sensitized with distinct dyes (SQ2 and D131). One cell responds mainly to blue light, the other to red. The voltage difference generated under light exposure is measured. Tests involved monochromatic light pulses (300–750 nm), synaptic behavior via paired-pulse experiments, classification of light pulse sequences, and motion recognition through color-coded human actions.
• Results: The device showed wavelength-dependent bipolar responses with 10 nm resolution, synaptic behavior with facilitation indices far exceeding conventional artificial synapses, and successful logic operations (AND, OR, XOR). It classified six-bit input patterns (64 states) and achieved 82% accuracy in recognizing combined motion and color in human actions, with perfect color classification.
• Limitations: Performance varies with light intensity, affecting wavelength discrimination. Motion recognition accuracy is slightly lower than some previous devices due to lower voltage outputs. The experiments used simple motion patterns under controlled conditions; real-world applications will need advanced signal processing. Additional circuitry is required to convert analog signals into clear digital outputs.

Funding and Publication Details

This work was supported by the Japan Science and Technology Agency (JST) through grants JPMJFS2144 and JPMJSP2151. The authors declared no conflicts of interest.

The study appeared in Scientific Reports (volume 15, article 16488) in 2025 under the title “Polarity-tunable dye-sensitized optoelectronic artificial synapses for physical reservoir computing-based machine vision.”

This development represents a step toward artificial vision systems combining the efficiency of biological sight with the precision of electronics. As demand for energy-efficient sensing and computing rises, such bio-inspired devices offer promising routes to smarter, low-power machine vision.