Get 3D object shape using neural implicit with self-supervision on surface normals with SN-NIR
Get 3D object shape using neural implicit with self-supervision on surface normals with SN-NIR
Normal-guided Detail-Preserving Neural Implicit Functions for High-Fidelity 3D Surface Reconstruction
arXiv paper abstract https://arxiv.org/abs/2406.04861
arXiv PDF paper https://arxiv.org/pdf/2406.04861
GitHub https://github.com/sn-nir/sn-nir
Project page https://sn-nir.github.io
Neural implicit representations … a powerful paradigm for 3D reconstruction … fail to capture fine geometric details and thin structures, especially … where only sparse RGB views
… hypothesize … methods … rely on 0-order differential properties, i.e. the 3D surface points and their projections, as supervisory signals. Such properties … do not capture the local 3D geometry around the points and also ignore the interactions between points.
… training neural representations with first-order differential properties, i.e. surface normals, leads to … accurate 3D surface reconstruction even … two RGB (front and back) images
… first compute the approximate surface normals in the image space using the gradient of the depth maps … using … monocular depth estimator such as Depth Anything model.
An implicit surface regressor is then trained using a loss function that enforces the first-order differential properties of the regressed surface to match those … from Depth Anything.
… show that the proposed method achieves an unprecedented level of reconstruction accuracy even when using as few as two RGB views …
Stay up to date. Subscribe to my posts https://morrislee1234.wixsite.com/website/contact
Web site with my other posts by category https://morrislee1234.wixsite.com/website
LinkedIn https://www.linkedin.com/in/morris-lee-47877b7b
