A new photonic chip achieves image processing at 25 million frames per second with high energy efficiency, promising major advances in real-time imaging and AI applications.
Key Details
- 1Vertically integrated photonic chip ('Gezhi') processes 25 million frames per second.
- 2Uses a mutually incoherent VCSEL array and diffractive neural network (DNN) for computation.
- 3Achieved 98.6% accuracy on MNIST image classification benchmark.
- 4Consumes only 3.52 aJ/μm² of optical energy per frame, outperforming electronic accelerators.
- 5Capable of versatile tasks like edge extraction and denoising.
- 6Can be scaled to larger distributed optical computing with tens of thousands of units.
Why It Matters
This technology could enable real-time, low-power AI implementation for high-throughput imaging tasks, which may dramatically enhance the speed and efficiency of image analysis in radiology and medical imaging fields. Its scalability and superior energy efficiency present opportunities for deploying advanced AI in areas such as autonomous diagnostics and smart healthcare.
Source
EurekAlert
Related News
•EurekAlert
AI-Powered OCT Enables Rapid 'Optical Biopsy' for Early Endometrial Cancer Detection
A team at Washington University has developed a catheter-based 3D OCT system with AI to quickly and noninvasively detect early endometrial cancers.
•EurekAlert
AI Clinical Reasoning in Diagnostics and Digital Fatigue in Healthcare
Recent JMIR features explore large language models in clinical diagnostics and digital fatigue among healthcare professionals.
•EurekAlert
KAIST, MIT, Microsoft Develop Efficient AI Image Upsampling for Robotics
KAIST, MIT, and Microsoft have created 'Upsample Anything,' a training-free AI method to restore high-resolution visual data from compressed images with up to 16x improved GPU memory efficiency.