HongwenZhang/PyMAF

[ICCV 2021, Oral] PyMAF: 3D Human Pose and Shape Regression with Pyramidal Mesh Alignment Feedback Loop

What it solves

PyMAF and PyMAF-X address the challenge of accurately regressing 3D human pose and shape from single monocular images or videos. They aim to improve the alignment and precision of the reconstructed 3D human mesh compared to the 2D image input, specifically focusing on full-body model regression for PyMAF-X.

How it works

The project utilizes a Pyramidal Mesh Alignment Feedback Loop to refine the 3D human mesh. It leverages the SMPL and SMPL-X body models to represent human shape and pose. The system can process both static images and videos (supporting single or multiple people), using pre-trained models to map image features to 3D mesh parameters.

Who it’s for

This tool is designed for researchers and developers working in computer vision, 3D human reconstruction, and human pose estimation.

Highlights

  • Full-body support: PyMAF-X extends the capability to well-aligned full-body model regression.
  • Versatile input: Supports both image and video inputs, including scenes with multiple people.
  • High performance: Ranked #2 on the AGORA Test (SMPL-X Whole-Body) leaderboard.
  • Feedback loop: Employs a pyramidal mesh alignment strategy to improve reconstruction accuracy.

Related

  • Project
    RizwanMunawar/yolov7-object-tracking

    A computer vision project that integrates YOLOv7 and YOLOv8 for real-time object detection and tracking across video streams from various sources.

  • Project
    chenwei-zhao/captcha-recognizer

    A deep learning-based library for identifying the gap coordinates and confidence levels in slider CAPTCHAs.

  • Project
    HZAI-ZJNU/Mamba-YOLO

    Mamba YOLO is an object detection baseline that replaces traditional architectures with State Space Models to provide an efficient alternative for real-time detection.

  • Project
    azxcvn/mpv-android-anime4k

    An Android video player based on libmpv that uses the Anime4K algorithm to provide real-time super-resolution upscaling for anime and animation videos.