The paper provides a comprehensive review of recent advancements in deep learning-based object detection methods for optical remote sensing images, highlighting the limitations of existing datasets and proposing a new benchmark dataset named DIOR. The DIOR dataset consists of 23,463 images and 192,472 instances, covering 20 object classes, and is characterized by its large scale, diverse object sizes, rich image variations, and high inter-class similarity and intra-class diversity. The paper also evaluates several state-of-the-art object detection methods on the DIOR dataset to establish a baseline for future research. Key contributions include a comprehensive survey of deep learning-based object detection methods, the creation of the DIOR dataset, and performance benchmarking on this dataset.The paper provides a comprehensive review of recent advancements in deep learning-based object detection methods for optical remote sensing images, highlighting the limitations of existing datasets and proposing a new benchmark dataset named DIOR. The DIOR dataset consists of 23,463 images and 192,472 instances, covering 20 object classes, and is characterized by its large scale, diverse object sizes, rich image variations, and high inter-class similarity and intra-class diversity. The paper also evaluates several state-of-the-art object detection methods on the DIOR dataset to establish a baseline for future research. Key contributions include a comprehensive survey of deep learning-based object detection methods, the creation of the DIOR dataset, and performance benchmarking on this dataset.