Journal: Advanced Science

Article Title: All‐Inorganic Perovskite Quantum‐Dot Optical Neuromorphic Synapses for Near‐Sensor Colored Image Recognition

doi: 10.1002/advs.202409933

Figure Lengend Snippet: Implementation of near‐sensor computation utilizing the CsPbBr 3 QD‐based ONS device to emulate the human visual system. a) Schematic of the neuromorphic vision system, comprising two nominally identical units (one serving as a photodetector and the other as an RRAM), seamlessly integrating to emulate colored image recognition in the human brain, and that of the block diagram illustrating the image recognition process within the 1P‐1R (near‐sensor) computing architecture. b,c) Schematics (top panels) elucidating ion migrations in the ONS and the RRAM devices, respectively, under applying either positive or negative electrical polarity to them. Their corresponding energy band diagrams are also plotted in the bottom panels of these figures. d) Equivalent circuit schematics illustrating the operational functionality of each unit constituted in the ONS device, under either positive (top panel) or negative (bottom panel) electrical polarity.

Article Snippet: Lee , All‐Inorganic Perovskite Quantum‐Dot Optical Neuromorphic Synapses for Near‐Sensor Colored Image Recognition .

Techniques: Blocking Assay

Journal: Advanced Science

Article Title: All‐Inorganic Perovskite Quantum‐Dot Optical Neuromorphic Synapses for Near‐Sensor Colored Image Recognition

doi: 10.1002/advs.202409933

Figure Lengend Snippet: Colored and color‐mixed image recognition by using the CsPbBr 3 QD‐based ONS device. a) Schematic illustration of the multiply‐accumulate operation of the neuromorphic encoding process using a 2D 28 × 28 array composed of the ONS devices. Corresponding matrix‐vector multiplication (bottom panel) is depicted with parameters of input optical signal V i (UV,B,G), synaptic weight W ij (UV,B,G), and output current signal I j (UV,B,G). b) Single‐layer artificial neural network illustrating the computational processes of using our ONS devices for colored image recognition. c) Recognition accuracy of the UV/B/G colored MNIST handwritten digits as a function of the number of training epochs. Insets: Confusion matrices of classification results for UV/B/G colored MNIST handwritten digits after 700 training epochs. d) Reconstructed mapping images of synaptic weights in accordance with ultraviolet (upper panel), blue (middle panel), and green (bottom panel) MNIST patterns after 700 training epochs. e) Encoded image (upper left) consisting of MNIST handwritten digits mixed with UV/B/G colored numerals, including 5 (UV), 3 (blue), and 1 (green), and intentionally introduced background noise. Classification results, obtained by feeding the encoded image into the trained neural network across various training epochs of 1, 10, 50, and 200 times, demonstrate distinct recognition capability for the MNIST handwritten digits in UV, blue, and green colors.

Article Snippet: Lee , All‐Inorganic Perovskite Quantum‐Dot Optical Neuromorphic Synapses for Near‐Sensor Colored Image Recognition .

Techniques: Plasmid Preparation