Skip to content

Latest commit

 

History

History
98 lines (67 loc) · 2.83 KB

README.md

File metadata and controls

98 lines (67 loc) · 2.83 KB
Raw

Single View to 3D README

Configuration

Update the data_location.py file with the root location of your dataset to match your system's directory structure.

Execution Instructions

Execute the following commands in your terminal. If necessary, replace python with the version specific to your system (e.g., python3).

Exploring Loss Functions

1.1 Voxel Representation

python fit_data.py --type 'vox'

1.2 Point Representation

python fit_data.py --type 'point'

1.3 Mesh Representation

python fit_data.py --type 'mesh'

2. Reconstructing 3D from a Single View

For training the model and reconstructing 3D from a single view, use the following commands:

2.1 Voxel-Based Model

  • This will create "checkpoint_vox.pth" in the root directory.
python train_model.py --type 'vox'
  • Evaluate the model with:
python eval_model.py --type 'vox' --load_checkpoint
  • Change the step parameter to select specific predictions to save.
  • Predictions and ground truths will be saved in "Visualization/Vis_vox".
Ground Truth Predicted
Ground Truth Predicted

2.2 Point-Based Model

  • This will create "checkpoint_point.pth" in the root directory.
python train_model.py --type 'point'
  • Evaluate the model with:
python eval_model.py --type 'point' --load_checkpoint
  • Predictions and ground truths will be saved in "Visualization/Vis_point".
Ground Truth Predicted
Ground Truth Predicted

2.3 Mesh-Based Model

  • This will create "checkpoint_mesh.pth" in the root directory.
python train_model.py --type 'mesh'
  • Evaluate the model with:
python eval_model.py --type 'mesh' --load_checkpoint
  • Predictions and ground truths will be saved in "Visualization/Vis_mesh".
Ground Truth Predicted
Ground Truth Predicted

After each evaluation, the threshold vs F1 score plot is saved in the "Visualization/Evaluation" directory.