3D Object Reassembly

Group 19: Mathias Vogel, Julia Bazińska, Katarzyna Krasnopolska, Davide Staub Project as a part of 3D Vision course.

Quickstart

Tested on Ubuntu 18.06 with Python 3.7.5.

pip install -r requirements.txt
cd full_pipeline
python3 reassemble_object.py --object_dir ./example_data/cube_10_seed_0

You should see first a visualization of a scrambled cube and then another one reassembled using ground truth data. In order to use the network predictions for keypoint matching, use the following command:

python3 reassemble_object.py --object_dir ./example_data/cube_10_seed_0 --use_predictions

Project structure

• evaluation – contains the scripts for evaluating keypoints and plotting the results.
• full_pipeline – reassemble an object from its shards, including all necessary intermediate steps.
• keypoints_and_descriptors – calculations of keypoints and descriptors.
• neural_network – build and train a neural network and use it for predictions.
• object_fracturing – fracture objects with blender and preprocess the objects for the neural network.
• object_reassembly – load preprocessed object fragments and reassemble either using ground truth matches or predictions from the network.

Troubleshooting

In case you encounter an error TypeError: 'exclusive' is an unknown keyword argument, then execute this with the appropriate path:

sed -i 's/QtWidgets.QActionGroup(self.window, exclusive=True)/QtWidgets.QActionGroup(self.window)/' $PATH_TO_YOUR_VIRTUAL_ENV/lib/python3.7/site-packages/compas_view2/app/app.py