3DVisLab 3D vision group at CSE@IITJ and CVIT@IIITH

4DHumans: Temporally Consistent Dynamic Human Capture


Kunwar Maheep Singh, Astitva Srivastava and Avinash Sharma



Our multi-view capture system


About

4DHumans dataset aims to capture multiple temporally consistent sequences of humans performing different actions. The sequences are expected to be roughly 1 minute long. The data is captured by using a multi-view setup involving multiple calibrated Microsoft Kinect cameras. Further, the sequences are post processed using the pipeline proposed in Fusion4D. We plan to make the dataset publicly available soon.

Dataset

Coming Soon!

Contact