ORIDa: Object-centric Real-world Image Composition Dataset

Figure 1. Overview of ORIDa. ORIDa contains 200 unique objects and over 30,000 real-captured images, including factual-counterfactual (F-CF) sets and factual-only (F-Only) images. F-CF sets consist of five images: one background-only and four with the object in different positions. F-Only images capture objects in diverse scenes, enhancing the diversity of the dataset for object reposition tasks.