# zhyever/PatchFusion

[CVPR 2024] An End-to-End Tile-Based Framework for High-Resolution Monocular Metric Depth Estimation

Repository: https://github.com/zhyever/PatchFusion
Canonical: https://ross.abutalabs.com/products/patchfusion
Homepage: https://zhyever.github.io/PatchFusion/
Language: Python
License: MIT
License Family: permissive
Topics: depth-estimation, high-resolution
Last push: 2026-02-21T11:27:16+00:00

## Health v2 (maintenance only)
Score: 57/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 68, release rhythm 35, longevity 71
- inputs: {"age_days": 1003, "days_push": 193, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1026, forks 64 (observed 2026-08-28T04:03:16.712561+00:00)

## What it is
PatchFusion is a CVPR 2024 end-to-end tile-based framework for high-resolution monocular metric depth estimation from single images. It fuses globally-consistent coarse depth predictions with finer tiled predictions and works on top of base models like ZoeDepth and Depth-Anything.

## Use cases
- estimate metric depth maps from high-resolution images
- run monocular depth estimation on 4K photos
- improve ZoeDepth or Depth-Anything depth predictions on large images
- train a depth estimation model on custom datasets
- generate depth maps for depth-guided image generation

## When to choose
- you need metric (not just relative) depth from single images at high resolution
- existing depth models lose detail or consistency on large images
- you want a research-grade, model-agnostic tiling framework with training code

## When to avoid
- you only need fast real-time depth on low-resolution video
- you need a production-ready plug-and-play API rather than a research codebase
- you lack a GPU, since inference is compute-intensive

## Facets
- artifact type: library
- maturity: active
- function: computer-vision, image-processing, machine-learning, deep-learning
- domain: computer-vision, machine-learning, image-processing, artificial-intelligence
- platform: python, cross-platform
- tags: depth-estimation, monocular-depth, metric-depth, tile-based-inference, high-resolution, cvpr-2024, pytorch, research-code, gpu, linux

## Member repositories
- zhyever/PatchFusion (main) score 57

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:16.712561+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:08:01.038489+00:00, confidence not recorded.
  - readme: https://github.com/zhyever/PatchFusion (fetched 2026-08-28T04:03:16.712561+00:00, sha 8378b16ff2f2)
  - homepage: https://zhyever.github.io/PatchFusion/ (fetched 2026-08-29T13:08:29.841448+00:00, sha c82cdffe16ea)
- Data as of 2026-08-30T08:39:29.467469+00:00.