🎯 Depth Anything V2 - Image to Point Cloud

Generate metric point clouds from single images using Depth Anything V2.

---

🚀 Quick Setup (One Command)

git clone https://github.com/CoderSATTY/image-to-pcd.git
cd image-to-pcd
chmod +x setup.sh && ./setup.sh

This downloads the model weights (~1.3GB) and sets up all dependencies.

---

📦 Usage

python depth_to_pointcloud.py \
  --encoder vitl \
  --load-from checkpoints/depth_anything_v2_metric_hypersim_vitl.pth \
  --max-depth 20 \
  --img-path ./input \
  --outdir ./output

Arguments:

Argument	Description	Default
--encoder	Model size: vits, vitb, vitl	vitl
--load-from	Path to model checkpoint	Required
--max-depth	Max depth in meters (20=indoor, 80=outdoor)	20
--img-path	Input image(s) path	Required
--outdir	Output directory for .ply files	Required
--calibration	Camera calibration file (.npz)	Optional

---

📁 Project Structure

image-to-pcd/
├── setup.sh                    # Run this first!
├── depth_to_pointcloud.py      # Main script
├── checkpoints/                # Model weights (auto-downloaded)
│   └── depth_anything_v2_metric_hypersim_vitl.pth
├── depth_anything_v2/          # Model code (auto-cloned)
├── input/                      # Place your images here
└── output/                     # Point clouds saved here

---

🔗 Model Download Links (Manual)

If you prefer manual download:

Model	Size	Download
Large (recommended)	335M	depth_anything_v2_metric_hypersim_vitl.pth
Base	97.5M	depth_anything_v2_metric_hypersim_vitb.pth
Small	24.8M	depth_anything_v2_metric_hypersim_vits.pth

Place in checkpoints/ directory.

---

🙏 Acknowledgements

• Depth Anything V2 - NeurIPS 2024