LiDAR Resolution Recovery via Foundation-Model-Guided Diffusion
Using a pretrained Stable Diffusion model as the backbone, we fine-tune a LiDAR-conditioned depth model with pseudo-depth targets from a 2D foundation model.
Key points
- High-beam-count LiDAR sensors are costly, yet many perception pipelines require dense angular sampling.
- Our model yields its largest advantage in very sparse regimes, achieving a $δ{1.25}$ accuracy of $66.8$% from $4$-beam input where scattered interpolation reaches only $45.1$%.
- A class-stratified error breakdown further reveals that planar surfaces recover first while objects introducing depth discontinuities degrade earliest.
- Together, these results quantify the recovery/resolution trade-off for foundation-model-guided LiDAR enhancement.
Sources (1)
- [1]LiDAR Resolution Recovery via Foundation-Model-Guided DiffusionarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 6, 04:20 PM
Using a pretrained Stable Diffusion model as the backbone, we fine-tune a LiDAR-conditioned depth model with pseudo-depth targets from a 2D foundation model.
High-beam-count LiDAR sensors are costly, yet many perception pipelines require dense angular sampling.
Extractive summary: sentences quoted from the sources.