Dataset_Synthesizer

Why we included this project

When you train object detectors or pose estimators, the bottleneck is often the data: real labeled images are hard to come by, especially when you need coverage across lighting conditions and camera angles. NDDS, NVIDIA's Unreal Engine-based dataset synthesizer, addresses this by generating photorealistic synthetic scenes and exporting them with labels automatically. Domain randomization varies textures, lighting, and object placement so models trained on the output generalize instead of overfitting to a single setup. The ground-truth annotations, including segmentation masks and 6-DoF pose labels, are exact by construction, which makes the tool attractive for robotics and industrial vision work. It runs as a C++/C# pipeline with an Unreal plugin, so it's a desktop workflow rather than a Python library; teams already working in Unreal will find it the most natural fit.