Adaptive synaptic phototransistor using MoS<sub>2</sub>/pV3D3 heterostructure for effective image segmentation
발표자
()
초록
내용
Neuronal systems-inspired optoelectronic devices are key to the advancement of machine vision technologies, which are critical for object recognition tasks. However, the dynamic nature of real-world lighting presents serious difficulties for currently used image processing methods. We present a novel phototransistor system that adapts to different lighting scenarios, resulting in enhancement in image segmentation efficiency. Our devices operate in regulatory modes that emulate neuronal regulatory mechanisms within the MoS2/pV3D3 heterostructure through multi-gating effects. The multi-gating functionality integrates segmented image componenets into a coherent visual representation without requiring numerous exposures or intensive post-processing. The performance evaluations using the Deeplab v3+ model with an optimized CamVid dataset demonstrate substantial enhancements in accuracy and Intersection over Union metrics, surpassing traditional methods.