What type of study is this?

This is a Quantitative Study study.

September 12, 2025Machine Learning Science and Technology0 citationsOpen Access

Self-Learning Framework for Unsupervised Depth Estimation on Non-Lambertian Surfaces

Physics-inspired self-learning framework for unsupervised depth estimation on non-Lambertian surfaces

Authors

This framework uncovers depth in images on non-Lambertian surfaces, suggesting improved accuracy and robustness.

The proposed framework improves depth estimation accuracy by addressing non-Lambertian surfaces and ghosting artifacts.
Achieving average improvements of 9.29% and 2.86% on Sq Rel and RMSE metrics respectively, the model shows significant performance gains.
Utilizing depth consistency loss and a Multi-Path Transformer, the approach combines photometric assumptions and game-theoretic strategies.
The method also displays strong zero-shot generalization capabilities, indicating its versatility across different datasets.

Ask AI

Helpful

Bookmark

View Full Paper

Li et al. (2025) studied this question.

Ask AI

Helpful

Bookmark

View Full Paper

September 12, 2025Machine Learning Science and Technology0 citationsOpen Access

Physics-inspired self-learning framework for unsupervised depth estimation on non-Lambertian surfaces

Authors

This framework uncovers depth in images on non-Lambertian surfaces, suggesting improved accuracy and robustness.

The proposed framework improves depth estimation accuracy by addressing non-Lambertian surfaces and ghosting artifacts.
Achieving average improvements of 9.29% and 2.86% on Sq Rel and RMSE metrics respectively, the model shows significant performance gains.
Utilizing depth consistency loss and a Multi-Path Transformer, the approach combines photometric assumptions and game-theoretic strategies.
The method also displays strong zero-shot generalization capabilities, indicating its versatility across different datasets.

Ask AI

Helpful

Bookmark

View Full Paper

Li et al. (2025) studied this question.

Ask AI

Helpful

Bookmark

View Full Paper