# shunsukesaito/PIFu

This repository contains the code for the paper "PIFu: Pixel-Aligned Implicit Function for High-Resolution Clothed Human Digitization"

Repository: https://github.com/shunsukesaito/PIFu
Canonical: https://ross.abutalabs.com/products/pifu
Homepage: https://shunsukesaito.github.io/PIFu/
Language: Python
License: NOASSERTION
License Family: other
Topics: pifu, pytorch, iccv2019, fashion, human, 3d-reconstruction, arvr, 3d-printing, geometry-processing, computer-graphics
Last push: 2023-11-24T17:55:31+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 2670, "days_push": 1013, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1817, forks 346 (observed 2026-08-28T04:05:40.586468+00:00)

## What it is
PyTorch implementation of PIFu (Pixel-Aligned Implicit Function), a deep learning method that reconstructs high-resolution 3D surfaces and textures of clothed humans from single or multiple images. It is the official research codebase for the ICCV 2019 paper, including test, training, and synthetic training-data generation pipelines.

## Use cases
- reconstruct a 3d human model from a single photo
- generate textured 3d meshes of clothed people from images
- digitize clothing wrinkles, hairstyles, and accessories into 3d geometry
- create 3d printable models of a person from a picture
- train a neural network to infer 3d human shape from 2d images
- build ar/vr avatar assets from real-world photos

## When to choose
- you need research-grade code for pixel-aligned implicit surface reconstruction of clothed humans
- you want to reproduce or build on the ICCV 2019 PIFu paper results
- you need high-resolution 3D human reconstruction with texture that handles arbitrary topology and unseen regions
- you want to generate your own training data with headless GPU rendering via EGL

## When to avoid
- you need a polished end-user application with a GUI rather than research code requiring Python and PyTorch setup
- you need real-time or production-supported 3D human reconstruction
- you want to reconstruct generic objects rather than human subjects
- you require actively maintained software with current framework versions (code targets older PyTorch and CUDA)

## Facets
- artifact type: library
- maturity: maintenance
- function: deep-learning, computer-vision, image-processing, machine-learning, graphics, data-generation
- domain: computer-vision, deep-learning, machine-learning, graphics, image-processing
- platform: python, cross-platform, windows
- tags: 3d-reconstruction, human-digitization, implicit-function, pytorch, single-image-3d, mesh-generation, texture-reconstruction, iccv2019, 3d-printing, arvr, research-code, geometry-processing, gpu, linux

## Member repositories
- shunsukesaito/PIFu (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:40.586468+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 2026-08-30T03:20:34.020773+00:00, confidence not recorded.
  - readme: https://github.com/shunsukesaito/PIFu (fetched 2026-08-28T04:05:40.586468+00:00, sha 9c6994c159e3)
  - homepage: https://shunsukesaito.github.io/PIFu/ (fetched 2026-08-29T10:59:34.340946+00:00, sha 528490308c1f)
- Data as of 2026-08-30T08:39:29.467469+00:00.