Skip to content

joaocmd/Inverse-Algorithmic-Design

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

57 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Inverse-Algorithmic-Design

Tool to convert floor plan images to code. To run, go to reconstruction, install the necessary Julia packages, and run julia InverseAlgorithmicDesign.jl, which will the options to use the tool.

Recognition

The Python code with the PyTorch models and vectorization mechanisms. Some notebooks are also useful, such as debug-brute-force.ipynb, which was used to visualize the vectorization algorithm. The qualitative-results.ipynb notebook was used to generate results for the qualitative evaluation of the system. The code to generate images after the recognition process is complete on that notebook, not on recognition.py.

The models' checkpoints to run this component are available here. The segmentation model model_best_val_loss_var.pkl is from CubiCasa5K.

The code in Recognition/raster_to_vector is adapted from the raster-to-vector original repository.

Reconstruction

Julia program that uses Conda.jl to install the Python packages and calls the recognition component with PyCall.jl. Then, $x$ and $y$ components for points are generated through hierarchical clustering and, finally, a new Julia program is written.

About

Tool to create Julia programs from floor plan images (MCS thesis) .

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published