Scale-Aware Monocular Depth Prediction for SLAM in Canonical Space

The full thesis is available as a PDF. Download or view it here. The code will become public later.

Overview of the results

The Scale-Aware method improves SLAM optimization by incorporating scale constraints into the Bundle Adjustment optimization step.

ScanNet Absolute Trajectory Error (ATE) in [m]

---

Citation

If you find my thesis useful in your research, please consider citing:

@thesis{Petropoulakis2020,
    author      = {Petropoulakis Panagiotis, S. B. Laina, S. Schaefer, J. Jung, and S. Leutenegger},
    title       = {Scale-Aware Monocular Depth Prediction for SLAM in Canonical Space},
    type        = {mscthesis},
    url         = {https://github.com/PetropoulakisPanagiotis/thesis/},
    institution = {TUM School of Computation, Information and Technology},
    year        = {2024},
}