Artistic application of monocular depth estimation – Vathography
retained text
Hello everyone, Happy to introduce Vathography, an art practice that turns depth maps estimated from a single photograph into artworks, as part of a photographic creative process. The monocular depth estimation model: Marigold V2 (Apache-2.0): [https://huggingface.co/spaces/huawei-bayerlab/marigold-v2-web](https://huggingface.co/spaces/huawei-bayerlab/marigold-v2-web) Paper: [https://arxiv.org/abs/2609.08084](https://arxiv.org/abs/2609.08084) To estimate depth maps from high-resolution photographs, I have developed a tiled ComfyUI node you can try here: [https://github.com/vathography/ComfyUI-Marigold-V2-Tiled-Depth-Vathography](https://github.com/vathography/ComfyUI-Marigold-V2-Tiled-Depth-Vathography) Open source (AGPL-3.0); the depth maps it produces are free to use, including for commercial work. The approach used to make high resolution work: one global pass (up to 2048 px) for the overall shape, 1536-px tiles for the detail, each tile aligned to the global estimate, then merged with a Laplacian pyramid. Would you recommend a different approach? Feedback and suggestions on how to improve it are welcome. Vathography explained 9min: [https://youtu.be/DVYgwRaTKkA](https://youtu.be/DVYgwRaTKkA) The practice uses free tools and is open to everyone. If you try it, I'm curious to see what you create
document fetched 2026-10-23 00:24 · canonical https://reddit.com/r/computervision/comments/1xcs1nh/artistic_application_of_monocular_depth