Enhancing Drone Vision with Synthetic Data Generation

Synclair Vision needed a more efficient and cost-effective way to gather the large volumes of labeled aerial imagery required to train their drone-based camera system—without relying on endless real-world flights. Modulai’s synthetic data pipeline automatically generates realistic aerial images and labeled data.

The Situation

Synclair Vision is developing an advanced drone-based camera system but struggles to collect labeled data to train the underlying vision models. Drone flights are time-consuming and it’s difficult to obtain data reflecting all conditions the system may face. They needed a more efficient and cost-effective way to gather the large volumes of labeled aerial imagery required to train their model. With synthetically-generated data, it becomes possible to generate large datasets for training these computer vision models.

The Solution

Modulai developed a synthetic data pipeline for Unreal Engine that artificially generates virtual scenes that mimic real-world environments. Given an outdoor scene, the system samples plausible camera positions and orientations based on a specified distribution. Objects of interest are then dynamically placed within the camera’s field of view, with randomized orientations. The pipeline captures both RGB images and corresponding segmentation masks. These masks are subsequently post-processed to extract bounding boxes around the objects

The Result

By automating data collection in a virtual environment, Synclair Vision greatly reduced reliance on costly, time-consuming drone flights. Thousands of labeled images can now be generated rapidly, covering a wide range of scenarios and object placements. Novel object classes may be easily added merely by importing new 3D assets.

ML Tools

Unreal Engine 5.4: 3D environment creation and model rendering.

UnrealCV: An open-source library providing Python interoperability with Unreal Engine and enabling automated capture of segmentation masks.

Python: The language of choice for pipelining, projection computations, and post-processing.: Manages camera setups, object placement, and labeling pipelines