Improving neural implicit surfaces geometry with patch warping

Neural implicit surfaces have become an important technique for multi-view 3D reconstruction but their accuracy remains limited. In this paper, we argue that this comes from the difficulty to learn and render high frequency textures with neural networks. We thus propose to add to the standard neural rendering optimization a direct photo-consistency term across the different views. Intuitively, we optimize the implicit geometry so that it warps views on each other in a consistent way. We demonstrate that two elements are key to the success of such an approach: (i) warping entire patches, using the predicted occupancy and normals of the 3D points along each ray, and measuring their similarity with a robust structural similarity (SSIM); (ii) handling visibility and occlusion in such a way that incorrect warps are not given too much importance while encouraging a reconstruction as complete as possible. We evaluate our approach, dubbed NeuralWarp, on the standard DTU and EPFL benchmarks and show it outperforms state of the art unsupervised implicit surfaces reconstructions by over 20% on both datasets. Our code is available at https://github.com/fdarmon/NeuralWarp

Automated summary

The paper proposes a method called NeuralWarp to improve the accuracy of neural implicit surface reconstruction by adding a direct photo-consistency term across different views. The approach optimizes the implicit geometry to achieve consistent view warping using predicted occupancy and normals, and measures their similarity using robust structural similarity. It also handles visibility and occlusion to encourage a complete reconstruction. Experimental results demonstrate that NeuralWarp outperforms state-of-the-art unsupervised implicit surface reconstructions by over 20% on standard benchmarks.