# gangweix/pixel-perfect-depth

[NeurIPS 2025] Pixel-Perfect Depth

Repository: https://github.com/gangweix/pixel-perfect-depth
Canonical: https://ross.abutalabs.com/products/pixel-perfect-depth
Homepage: https://pixel-perfect-depth.github.io/
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2026-02-13T00:02:46+00:00

## Health v2 (maintenance only)
Score: 49/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 67, release rhythm 35, longevity 33
- inputs: {"age_days": 464, "days_push": 202, "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 1064, forks 39 (observed 2026-08-28T04:03:26.340134+00:00)

## What it is
Pixel-Perfect Depth is a monocular depth estimation model based on pixel-space diffusion transformers that produces flying-pixel-free depth maps and point clouds. It includes model weights, training, inference, and evaluation code released alongside a NeurIPS 2025 paper.

## Use cases
- estimate depth maps from a single image
- generate point clouds from monocular depth without flying pixels
- run diffusion-based depth estimation on images
- evaluate monocular depth models on zero-shot benchmarks
- train a pixel-space diffusion depth model
- convert photos to 3D point clouds

## When to choose
- you need high-quality depth maps with clean edges for 3D reconstruction or point clouds
- you want a state-of-the-art generative monocular depth model with released weights
- you're researching diffusion transformers for dense prediction

## When to avoid
- you need real-time depth estimation on low-power hardware
- you want a simple lightweight discriminative depth model without GPU requirements
- you need a production-ready packaged application rather than research code

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, computer-vision, image-processing, llm-inference
- domain: computer-vision, machine-learning, deep-learning, artificial-intelligence
- platform: python
- tags: monocular-depth-estimation, diffusion-transformers, point-clouds, depth-estimation, research-code, neurips-2025, gpu, linux

## Member repositories
- gangweix/pixel-perfect-depth (main) score 49

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:26.340134+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 2026-08-30T06:56:04.090154+00:00, confidence not recorded.
  - readme: https://github.com/gangweix/pixel-perfect-depth (fetched 2026-08-28T04:03:26.340134+00:00, sha 16bf1223d972)
  - homepage: https://pixel-perfect-depth.github.io/ (fetched 2026-08-29T12:58:06.838061+00:00, sha d2ccde5afa10)
- Data as of 2026-08-30T08:39:29.467469+00:00.