How derkleineli/meshpad-release works
MeshPad is an interactive sketch-conditioned mesh creation and editing application, alongside code for training its addition and deletion models. The user workflow is to draw or erase sketch strokes, then request a mesh addition or deletion and inspect the updated mesh. The graph includes documented training and rendering capabilities; demo internals and unsampled model/engine wiring are not expanded beyond what the available excerpts support.
Main entry points: gradio_demo.py Gradio demo · meshpad_addition.py Add part · meshpad_deletion.py Delete part · train_addition.py Addition trainer
Components 19
Interactive editing
- Gradio demo
demo/gradio_demo.py - Sketch input
demo/gradio_demo.py - Add part
meshpad/models/meshpad_addition.py - Delete part
meshpad/models/meshpad_deletion.py
Mesh models
- Shared model base
meshpad/models/base.py - Addition ShapeOPT
meshpad/models/shape_opt_addition.py - Deletion ShapeOPT
meshpad/models/shape_opt_deletion.py
Model training
- Addition training
meshpad/engine_addition.py - Deletion training
meshpad/engine_deletion.py - Addition trainer
train_addition.py - Deletion trainer
train_deletion.py - OBJ dataset
meshpad/dataset/obj_folder_dataset.py - Training options
meshpad/utils/argsparser.py
Mesh representation
- Sketch renderer
meshpad/utils/pyrender_sketch.py - Addition preprocessing
meshpad/engine_addition.py - Deletion preprocessing
meshpad/engine_deletion.py - Mesh result
Other components
- Artist
- RADIO vision model
Connections 26
- Artist → Gradio demo: uses
- Gradio demo → Sketch input: edits
- Sketch input → Gradio demo: updates sketch
- Gradio demo → Add part: requests addition
- Gradio demo → Delete part: requests deletion
- Add part → Mesh result: generates mesh
- Delete part → Mesh result: edits mesh
- Addition trainer → Training options: parses options
- Deletion trainer → Training options: parses options
- Addition trainer → OBJ dataset: loads dataset
- Deletion trainer → OBJ dataset: loads dataset
- Addition trainer → Add part: initializes model
- Deletion trainer → Delete part: initializes model
- Addition trainer → Addition training: starts training
- Deletion trainer → Deletion training: starts training
- Addition training → Addition preprocessing: prepares batches
- Deletion training → Deletion preprocessing: prepares batches
- Addition preprocessing → Sketch renderer: renders sketches
- Deletion preprocessing → Sketch renderer: renders sketches
- Addition preprocessing → OBJ dataset: reads batches
- Deletion preprocessing → OBJ dataset: reads batches
- Add part → Shared model base: uses
- Delete part → Shared model base: uses
- Shared model base → RADIO vision model: extracts features
- Add part → Addition ShapeOPT: uses decoder
- Delete part → Deletion ShapeOPT: uses decoder