MonoTrap
收藏资源简介:
MonoTrap数据集由博洛尼亚大学的研究人员创建,专门用于测试光学幻觉对单目深度估计模型的影响。该数据集包含26个场景,每个场景都有相应的真实深度图,旨在模拟单目深度估计模型在面对复杂视觉幻觉时的表现。数据集的创建过程结合了现代图形引擎生成的高质量合成数据,确保了数据的多样性和复杂性。MonoTrap数据集主要应用于计算机视觉领域,特别是深度估计和立体匹配任务,旨在解决单目深度估计模型在复杂场景下的鲁棒性问题。
The MonoTrap dataset was created by researchers at the University of Bologna, specifically designed to test the impact of optical illusions on monocular depth estimation models. This dataset contains 26 scenes, each paired with a corresponding ground-truth depth map, aiming to simulate the performance of monocular depth estimation models when confronted with complex visual illusions. The creation process of the dataset integrates high-quality synthetic data generated by modern graphics engines, ensuring the diversity and complexity of the dataset. Primarily utilized in the field of computer vision, particularly for depth estimation and stereo matching tasks, the MonoTrap dataset is intended to address the robustness issues of monocular depth estimation models in complex scenes.

- 1Stereo Anywhere: Robust Zero-Shot Deep Stereo Matching Even Where Either Stereo or Mono Fail博洛尼亚大学 · 2024年



