Search papers, labs, and topics across Lattice.
Marigold-SSD, a novel single-step depth completion framework, is introduced to address the computational expense of diffusion-based methods by shifting the computational burden to finetuning. This late-fusion approach leverages strong diffusion priors for efficient and robust 3D perception, achieving significantly faster inference times. Evaluated on six benchmarks, Marigold-SSD demonstrates strong cross-domain generalization and zero-shot performance, effectively bridging the efficiency gap between diffusion and discriminative models.
Ditch the slow diffusion grind: Marigold-SSD delivers zero-shot depth completion in a single step, rivaling discriminative models in speed while retaining diffusion's accuracy.
We introduce Marigold-SSD, a single-step, late-fusion depth completion framework that leverages strong diffusion priors while eliminating the costly test-time optimization typically associated with diffusion-based methods. By shifting computational burden from inference to finetuning, our approach enables efficient and robust 3D perception under real-world latency constraints. Marigold-SSD achieves significantly faster inference with a training cost of only 4.5 GPU days. We evaluate our method across four indoor and two outdoor benchmarks, demonstrating strong cross-domain generalization and zero-shot performance compared to existing depth completion approaches. Our approach significantly narrows the efficiency gap between diffusion-based and discriminative models. Finally, we challenge common evaluation protocols by analyzing performance under varying input sparsity levels. Page: https://dtu-pas.github.io/marigold-ssd/