NoPe-NeRF: Optimising Neural Radiance Field with No Pose Prior
This work addresses a challenging issue in 3D scene reconstruction for computer vision applications, but it is incremental as it builds on existing methods for joint optimization of NeRF and camera poses.
The paper tackles the problem of training Neural Radiance Fields (NeRF) without pre-computed camera poses, especially under dramatic camera movement, by incorporating undistorted monocular depth priors and novel loss functions. The result is improved novel view rendering quality and pose estimation accuracy, as demonstrated on real-world indoor and outdoor scenes.
Training a Neural Radiance Field (NeRF) without pre-computed camera poses is challenging. Recent advances in this direction demonstrate the possibility of jointly optimising a NeRF and camera poses in forward-facing scenes. However, these methods still face difficulties during dramatic camera movement. We tackle this challenging problem by incorporating undistorted monocular depth priors. These priors are generated by correcting scale and shift parameters during training, with which we are then able to constrain the relative poses between consecutive frames. This constraint is achieved using our proposed novel loss functions. Experiments on real-world indoor and outdoor scenes show that our method can handle challenging camera trajectories and outperforms existing methods in terms of novel view rendering quality and pose estimation accuracy. Our project page is https://nope-nerf.active.vision.