Abstract: A temperate cyclone is an important weather system in the mid-latitude regions and has a profound impact on global weather and climate. The frontal system is a key structure of a temperate cyclone, and its accurate identification is crucial for improving weather forecast accuracy. In recent years, deep learning methods have been applied in the field of frontal automatic recognition, but research on frontal identification over the Eurasian continent has been limited, often focusing on detecting single-type fronts. Achieving simultaneous recognition and classification of cold and warm fronts still presents significant challenges. Therefore, this paper proposes a cold and warm front synchronous recognition method based on the DETR (Detection Transformer) model, called DETR-FRO. The method constructs RGB feature images using meteorological variables such as sea-level pressure, temperature advection at 850 hPa, temperature, and relative humidity, and trains the DETR model with an artificial cold and warm front label set to achieve automatic synchronous recognition and classification of cold and warm fronts. DETR-FRO improves the original DETR model’s transmission channel by incorporating multiple meteorological variables to construct RGB feature images, enhancing the ability to express frontal features, resulting in good recognition performance for both cold and warm fronts. Multi-dimensional validation results show that the method can simultaneously identify the cold and warm fronts of temperate cyclones over the Eurasian continent with high accuracy and stability. In the application to a typical snowstorm case, the DETR-FRO model accurately restores the cold front, warm front, and their evolution process, and the recognition results align well with dynamic and thermodynamic fields, further verifying the model"s generalization ability and meteorological interpretability. This improvement also provides new technical support for weather forecasting and the study of temperate cyclones.