A Simulation Dataset Generator for Object Detection Research

This paper proposes a method to address the issues of limited training data, difficulty in data acquisition, and high cost in the research and application of intelligent perception algorithms such as object detection. The simulation dataset generator is proposed to construct object detection simulation datasets based on the Unreal Engine, which can achieve low-cost and fast construction of simulation datasets for specific targets in specific scenarios. By building a 3D environment and targets of interest in the Unreal Engine, together with camera and IMU simulation sensors, we can efficiently generate multi-view simulation datasets. The simulation dataset generator has strong universality and can quickly construct datasets for different scenarios and targets. The experiment proves the effectiveness of the simulation dataset.