CL-MVSNet: Unsupervised Multi-view Stereo with Dual-level Contrastive Learning
This work addresses limitations in 3D reconstruction for computer vision applications, offering an incremental improvement over existing unsupervised methods.
The paper tackles the problem of inaccurate depth estimation in unsupervised multi-view stereo due to indistinguishable regions and view-dependent effects by proposing CL-MVSNet with dual-level contrastive learning and an L0.5 photometric consistency loss, achieving state-of-the-art performance on DTU and Tanks&Temples benchmarks and outperforming supervised methods without fine-tuning.
Unsupervised Multi-View Stereo (MVS) methods have achieved promising progress recently. However, previous methods primarily depend on the photometric consistency assumption, which may suffer from two limitations: indistinguishable regions and view-dependent effects, e.g., low-textured areas and reflections. To address these issues, in this paper, we propose a new dual-level contrastive learning approach, named CL-MVSNet. Specifically, our model integrates two contrastive branches into an unsupervised MVS framework to construct additional supervisory signals. On the one hand, we present an image-level contrastive branch to guide the model to acquire more context awareness, thus leading to more complete depth estimation in indistinguishable regions. On the other hand, we exploit a scene-level contrastive branch to boost the representation ability, improving robustness to view-dependent effects. Moreover, to recover more accurate 3D geometry, we introduce an L0.5 photometric consistency loss, which encourages the model to focus more on accurate points while mitigating the gradient penalty of undesirable ones. Extensive experiments on DTU and Tanks&Temples benchmarks demonstrate that our approach achieves state-of-the-art performance among all end-to-end unsupervised MVS frameworks and outperforms its supervised counterpart by a considerable margin without fine-tuning.