Christopher-Lim commited on
Commit
0d3e3a6
·
verified ·
1 Parent(s): c7c5534

Update README.md

Browse files

---

# ◈ GraphiContact ◈

GraphiContact is a pose-aware framework for **3D human mesh reconstruction** and **human-scene contact point prediction** from **monocular RGB images**. It is designed to improve contact perception and reconstruction robustness in complex environments by integrating pose-aware features with **SIMU (Single-Image Multi-Infer Uncertainty) modeling**.

This repository hosts the model weights associated with the paper:

**GraphiContact: Pose-aware Human-Scene Robust Contact Perception for Interactive Systems**
Accepted at **ICME 2026**.

## Highlights

- Monocular RGB-based 3D human mesh reconstruction
- Human-scene contact point prediction
- Pose-aware feature integration
- SIMU-based robustness modeling for challenging observations such as occlusion and visual perturbation

## Model Description

GraphiContact combines pose-aware human modeling with human-scene interaction reasoning to improve both contact prediction and 3D reconstruction quality. The method further introduces SIMU modeling to expose the model to perceptual variations during training, improving robustness in real-world scenarios.

## Intended Use

This model is intended for research use in:

- 3D human reconstruction
- Human-scene contact perception
- Interactive systems
- Embodied AI
- Human-centered visual understanding

## Training / Evaluation Context

According to the associated project repository, GraphiContact is evaluated with code, demos, and experiment scripts released for multiple benchmarks relevant to human reconstruction and contact perception.

## Usage

After downloading the checkpoint, load it with the GraphiContact codebase from the official repository:

```bash
git clone https://github.com/Aveiro-Lin/GraphiContact.git
cd GraphiContact

Files changed (1) hide show
  1. README.md +17 -3
README.md CHANGED
@@ -1,3 +1,17 @@
1
- ---
2
- license: mit
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: mit
3
+ language:
4
+ - en
5
+ metrics:
6
+ - recall
7
+ - precision
8
+ - f1
9
+ pipeline_tag: image-to-3d
10
+ tags:
11
+ - 3d-human-reconstruction
12
+ - human-scene-contact
13
+ - monocular-rgb
14
+ - mesh-reconstruction
15
+ - pose-aware
16
+ - icme-2026
17
+ ---