A differential perovskite hemispherical photodetector is introduced for intelligent imaging and location tracking. This device, with 8 differential pixels, enables high external quantum efficiency (-1000%) and low noise (10^-13 A Hz^-0.5), allowing for stable and large signal responses. By analyzing the differential light response of 8 pixels using computer algorithms, the device achieves colorful imaging and a computational spectral resolution of 4.7 nm in a low-cost, lensless geometry. Machine learning is used to mimic differential current signals under different biases, enabling dynamic tracking of objects in 3D space or 2D planes with color classification. The device integrates advanced functions for intelligent, low-cost photodetectors, essential for modern technology. It addresses the limitations of conventional imaging systems by using a single-pixel device to capture 2D images through Fourier transform and machine learning. The device's hemispherical structure provides wide-angle detection and enhanced responsivity to light intensity, wavelength, and object distance. Spray-coated perovskite films with supramolecular aggregates improve film quality and charge injection, leading to high EQE and device gain. The differential photodetector enables computational spectrometry with high resolution and color imaging through machine learning. It also supports location tracking by capturing light source position information. The device's hemispherical structure allows for differential signals to be captured, enabling 3D spatial tracking. The device's performance is validated through experiments, showing high accuracy in color classification and location tracking. The device's design and performance demonstrate its potential for future intelligent applications in artificial intelligence.A differential perovskite hemispherical photodetector is introduced for intelligent imaging and location tracking. This device, with 8 differential pixels, enables high external quantum efficiency (-1000%) and low noise (10^-13 A Hz^-0.5), allowing for stable and large signal responses. By analyzing the differential light response of 8 pixels using computer algorithms, the device achieves colorful imaging and a computational spectral resolution of 4.7 nm in a low-cost, lensless geometry. Machine learning is used to mimic differential current signals under different biases, enabling dynamic tracking of objects in 3D space or 2D planes with color classification. The device integrates advanced functions for intelligent, low-cost photodetectors, essential for modern technology. It addresses the limitations of conventional imaging systems by using a single-pixel device to capture 2D images through Fourier transform and machine learning. The device's hemispherical structure provides wide-angle detection and enhanced responsivity to light intensity, wavelength, and object distance. Spray-coated perovskite films with supramolecular aggregates improve film quality and charge injection, leading to high EQE and device gain. The differential photodetector enables computational spectrometry with high resolution and color imaging through machine learning. It also supports location tracking by capturing light source position information. The device's hemispherical structure allows for differential signals to be captured, enabling 3D spatial tracking. The device's performance is validated through experiments, showing high accuracy in color classification and location tracking. The device's design and performance demonstrate its potential for future intelligent applications in artificial intelligence.