Abstract: Internal solitary waves (ISWs) in the ocean usually carry significant amount of energy, posing serious safety hazards to offshore operations, making their real-time detection crucial. In recent years, the application of artificial intelligence (AI) technology has become increasingly widespread across various industries, with deep leaning-based object detection technologies maturing. This paper present an AI-based detection model for ISWs, developed using mooring observation data in combination with YOLOv5 deep leaining object detection algorithm. Firstly, the data are normalized, and different observational elements are fed into the network for iterative training. After the training, the optimal weight files are saved and used to evaluate the model. The evaluation results indicate that in the single-element detection scheme, the temperature detection model performs slightly better than the zonal flow rate model, while the meridional flow rate model is relatively less effective, with the maximum values of the comprehensive evaluation factor \varepsilon _\mathrmm\mathrma\mathrmx being 80.94, 78.74 and 72.02, respectively. In the detection scheme using combinations of temperature, zonal flow rate, and meridional flow rate, the three-element scheme outperforms the single- and two-element scheme, with \varepsilon _\mathrmm\mathrma\mathrmx reaching 82.48. This detection scheme demonstrates high accuracy and fast detection speed, meeting requirements for the real-time detection.