Fig. 1. Schematic of the architecture and performance of the OGPU named Gezhi. (IMAGE)
Caption
Fig. 1. Schematic of the architecture and performance of the OGPU named Gezhi. a. Conceptual diagram of the OGPU designed for high-throughput image processing, with each frame processed in 40 ns. b. Demonstration of the tasks performed by the OGPU including image classification and image processing (edge extraction and denoising). c. Architecture overview, featuring an addressable mutually incoherent VCSEL array for generating 2D signals, a MI-DNN for multi-channel processing, and detectors for output signal acquisition. The VCSELs illuminate the distinct regions of the MI-DNN enabling spatial weight multiplexing. d. Optical image of the OGPU and VCSEL array. Scale bar: 200 μm. e. Demonstration of the VCSEL array’s addressability, displaying the letters “U”, “S”, “S”, and “T”. Scale bar: 200 μm. f. Pulse signals generated by the VCSEL array, operating at different frequencies within a 1000 ns timeframe.
Credit
Min Gu et al.
Usage Restrictions
Credit must be given to the creator.
License
CC BY