3D Human Mesh Generator from Single Image

This project is a pipeline that detects a person in an image, segments them precisely, and reconstructs a 3D mesh from the cropped region. It integrates YOLO, SAM, OpenPose, and PIFuHD to generate high-fidelity 3D models from 2D images.

---

🔧 Pipeline Steps

1. Detection & Masking

   • Run main.py
   • Executes YOLO for detection
   • Applies SAM for precise segmentation
   • Uses OpenPose for keypoint detection
   • You will be prompted to save the cropped image → Save it to:

     pifuhd/openpose/example/media/
     

2. 3D Mesh Reconstruction

   • Navigate to pifuhd/
   • Run:

     python -m apps.simple_test
     

   • This generates the 3D mesh output using PIFuHD.

---

🧠 Models Used

• Yolov11 – Person detection from full image.
• SAM (Segment Anything Model) – Precise mask of detected person.
• OpenPose – Extracts pose landmarks to improve mesh accuracy.
• PIFuHD – Reconstructs 3D mesh from masked image input.

---

📦 Folder Structure

.
├── main.py                # Orchestrates YOLO → SAM → OpenPose
├── /pifuhd
│   ├── /openpose/example/media/  # Save masked crop here
│   └── apps/simple_test.py       # Run this for 3D mesh output

---

📷 Example Outputs

Stage	Output
1. YOLO Detection
2. SAM Masking
3. 3D Mesh (PIFuHD)

---

⚠️ Limitation

• Requires the user to manually save the cropped masked image at the specific location before mesh reconstruction.

---

🙌 Acknowledgements

This project integrates and builds upon the following repositories:

• 🔗 Ultralytics: https://github.com/ultralytics/ultralytics
• 🔗 SAM (Segment Anything): https://github.com/facebookresearch/segment-anything
• 🔗 PIFuHD: https://github.com/facebookresearch/pifuhd

---

🏁 Getting Started

# 1. Clone all required repos
git clone https://github.com/Faykar78/AI_powered_3d_model_generator.git
cd your-repo

# 2. Setup environments for each dependency (YOLO, SAM, OpenPose, PIFuHD)

# 3. Start detection pipeline
python main.py

# 4. Save cropped output manually as instructed

# 5. Run mesh generation
cd pifuhd
python -m apps.simple_test