Namespace:    svcl

PI: Nuno Vasconcelos
Institution: University of California, San Diego
Project description:

Data collection in the real world is very expensive. However, there are infinite sources of synthetic data from gaming environments, which are very easy/cheap to collect. We are exploring the impact of synthetic data on the training of real-world computer vision systems. This involves collecting a large amount of synthetic data from game engines, training vision models with this data, and measuring their performance in real-world computer vision tasks, e.g. object detection.

Software: Python, CUDA, PyTorch, TensorFlow, Caffe, NumPy, OpenCV

Back to namespaces