Embedded Files
A New Take on the Structure from Motion Pipeline
A New Take on the Structure from Motion Pipeline
Structure from Motion (SfM) is a pipeline that is found throughout modern-day vision-based perception algorithms. Whether this is the auto brake function in cars or something as trivial as the mapping capabilities within the iRobot Roomba, this process plays an enormous role in our everyday lives when it comes to autonomy. However, these algorithms are plagued with issues at it's current, especially when it comes to reliability.
A quick summary of what SplattingTAPIR has to offer.
Utilizing the Best Models
Utilizing the Best Models
SplattingTAPIR is a new twist on this pipeline that combines the best machine-learning algorithms. In collaboration with the Havard Computational Robotics Lab, Dr. Nguyen, and support from the Google TPU Research Cloud, models like Depth-Anything, TAPIR, and 3D Gaussian Splatting are researched to create a robust and efficient pipeline that outperforms classical approaches in large magnitudes.
Page updated
Google Sites
Report abuse