MV-SyDog: A Multi-View Synthetic Dog Dataset

Data Information

We introduce MV-SyDog, a synthetic dataset featuring multi-view videos of dogs. It includes 1K videos, each lasting 2 seconds with 2D/3D pose ground truth and depth maps. We enhanced the dataset's ground truth by extending the Unity Perception package to capture 3D kinematic motion sequences for each video. For added scene diversity, we included five different dog models with textures, representing various dog breeds and sizes. Six virtual cameras were positioned in a circular arrangement around the subject to produce the multiple viewpoints. To introduce further variation, 420 different high dynamic range images (HDRIs) were used for realistic lighting and backgrounds. Randomness was introduced in the dog's appearance and pose, HDRIs, and view settings for each video.

License Agreement

License file is in license.txt

1. All original images and associated data provided may be used for non-commercial research purposes only.
2. The source of the datasets must be acknowledged in all publications where they are used.
3. The data may not be redistributed.