BID-NeRF: RGB-D image pose estimation with inverted Neural Radiance Fields

We aim to improve the Inverted Neural Radiance Fields (iNeRF) algorithm which defines the image pose estimation problem as a NeRF based iterative linear optimization. NeRFs are novel neural space representation models that can synthesize photorealistic novel views of real-world scenes or objects. Our contributions are as follows: we extend the localization optimization objective with a depth-based loss function, we introduce a multi-image based loss function where a sequence of images with known relative poses are used without increasing the computational complexity, use a sequence of images with known relative poses and extend the sampling interval to encounter higher errors in the initial pose estimate, we omit hierarchical sampling during volumetric rendering, meaning only the coarse model is used for pose estimation, and we show that by extending the sampling interval convergence can be achieved even for higher initial pose estimate errors. With the proposed modifications the convergence speed is significantly improved, and the basin of convergence is substantially extended.