Harnessing the Potential of Advanced Large Vision Models to Enhance the Detection of Optoelectronic Imaging Signals

This study focuses on exploring the use of SAM (Segment Anything Model), an advanced visual foundation model, to enhance the detection of optoelectronic imaging signals. We fine-tuned the mask encoder of SAM and used the Electron Microscopy Dataset as the experimental dataset. To evaluate the effect, the U-net model was also used as a comparison benchmark. The experimental results show that the IoU metrics of SAM outperform those of U-net when only a small amount of data is available, demonstrating that the fine-tuned SAM has a unique advantage in recognizing photoelectric imaging signals.