<p>Although various models have been proposed for the purpose of oil spill monitoring, most of them should be categorized as image segmentation algorithms, which separate the image into several areas based on their different appearances. In contrast, the object detection models, which are commonly used for pattern recognition in computer graphics, have barely been applied to monitor oil spill. This study conducted experiments on oil spill monitoring using object detection models to clarify the existing questions on monitoring oil spills using such type of models. Specifically, the faster-RCNN model was optimized using a feature pyramid network (FPN) to detect the oil spills in both hyperspectral and synthetic aperture radar (SAR) images. The results indicated that the FPN-integrated faster-RCNN model can detect the oil spills in both hyperspectral and SAR images at reasonable accuracy. According to the results, the FPN network improved the precision, recall, and average precision (AP) by 0.1. For the hyperspectral data, the object detection model achieved higher precision, recall, and AP over 0.9 using the pseudo-colour images. For the SAR images, the object detection model achieved the highest accuracies using the polarization characteristic parameter of anisotropy. It is found that the object detection models are superior to the image segmentation models under the application situations where the binary determination on the existence of an oil spill is needed with the requirements of real-time processing and low false-alarm rate. Thus, it is expected that the object detection model could be applied to the daily surveillance of oil spills using airborne remote sensor. Moreover, they could be integrated with the image segmentation algorithms to reach the balance between model efficiency and the extraction of exact oil spill areas.</p>