Code for our MeshRunner project
In this work we tried to improve several aspects in MeshWalker:
- Improved walk generation
- Choose starting point by analyzing saliency
- Apply skips, jumps and ordering
- Improved walk representation
- Attention based information exchange between walks
The authors of this project are:
- Itay Levy
- Itamar Zimerman
- Amit Cohen
Prof. Amit Bermano - Guided us thorought the project
Alon Lahav - Help reproducing MeshWalker results
Adi Mesika - Helped with visualizations
- Apply Deep Learning directly on meshes
- Perform random walks on the mesh’s surface and feed it to an RNN
- Use the RNN’s hidden state to represent the mesh
- Perform classification (or segmentation) over the representation
Checkout our colab notebook for a demo of this project:
