Scan2Mesh

A non-official PyTorch implementation for Scan2Mesh: From Unstructured Range Scans to 3D Meshes paper on ShapeNet dataset

Installation

pip install -r requirements.txt

Download ShapeNet dataset and store it in data folder
Reduce the meshes in ShapeNet to at most 100 vertices using the notebook data/Generation Notebook.ipynb
Train the vertix prediction model using the following command:
```
python train_vertix.py --config=configs/vertix_train.json
```
Train the edge prediction model using the following command:
```
python train_edge.py --config=configs/edge_train.json
```
Train the face prediction model using the following command:
```
python train_face.py --config=configs/face_train.json
```
Use the visualization notebook Visualize Results.ipynb to visualize the trained model on ShapeNet meshes.

Name		Name	Last commit message	Last commit date
Latest commit History 25 Commits
.ipynb_checkpoints		.ipynb_checkpoints
configs		configs
data		data
inference		inference
model		model
training		training
util		util
.gitignore		.gitignore
README.md		README.md
Visualize Results.ipynb		Visualize Results.ipynb
evaluate.py		evaluate.py
generate_edge_meshes.py		generate_edge_meshes.py
main.py		main.py
requirements.txt		requirements.txt
train_edge.py		train_edge.py
train_face.py		train_face.py
train_vertix.py		train_vertix.py
visualizations.png		visualizations.png