Owing to its exceptional carrier mobility and unique topological features, Bi2Te3 emerges as an ideal material for crafting high-performance photodetectors. The synergistic integration of Bi2Te3 photodetectors with convolutional neural networks (CNNs) holds the potential to enhance feature extraction robustness and target recognition accuracy. This study focuses on the optoelectronic properties of the Bi2Te3/WS2 heterojunction and explores its application in Image recognition in combination with CNNs. By fabricating the Bi2Te3/WS2 photodetector, the dark current of the device has been effectively reduced, enabling photodetection within the visible-near-infrared wavelength range. The device has a detectivity of 8.39 × 108 cm·Hz1/2·W−1 and a response time of 185 μs/71 μs. Furthermore, the Bi2Te3/WS2 photodetector enables high-resolution image imaging. Leveraging CNNs, the image recognition accuracy attains 88 %. This research not only illuminates the construction of Bi2Te3 nanostructured optoelectronic heterojunction devices but also presents an innovative approach for image recognition applications.