Visible–near-infrared (VIS-NIR) dual-mode imaging can expand the human perception limit. However, the development of dual-mode image sensors is still challenging due to complex fabrication processes and readout circuit design. Here, we design a simple-structured (perovskite-Au/Si/Ag) dual-mode photodetector. This unique asymmetric electrode design allows the device to support two operating modes at zero bias. The device exhibits a detection range covering 400–1100 nm under top illumination mode, with a peak specific detectivity of up to 5.56×1013 Jones. Under bottom illumination mode, the device demonstrates pronounced narrowband NIR response characteristics. More importantly, we develop a dual-mode single-pixel imaging system based on this device, bypassing the fabrication processes of high-density array image sensors. The system exhibits excellent VIS-NIR dual-mode imaging results, effectively separating NIR and VIS information and enhancing infrared details in the fused images. Interestingly, we discover that the system can effectively suppress ringing artifacts, achieving infrared information perception at a low sampling rate, which can accelerate the imaging speed by ∼16 times (reduced from ∼3.2 s to ∼0.2 s). Our proposed dual-mode single-pixel imaging technology offers new means for material identification and intelligent perception.