1.  Install all dependencies

(1) To create a conda environment with all the required packages using conda run the following command:

conda env create -f resources/deformprior_env.yml

The above command creates a conda environment with the name deformprior.

(2) Compile external dependencies inside external directory by executing:
conda activate deformprior
./build_external.sh


(3) We use farthest point sampling (FPS) to downsample the input. To install the cuda implementation of FPS, you need to run

pip install pointnet2_ops_lib/.

python setup.py build_ext --inplace





2. Generate data for training & validation

./scripts/generate_data_deform4d_seq_animals.sh



3. Training

(1) forward deformation network

./scripts_train_val/train_forward.sh

(2) backward deformation network

./scripts_train_val/train_backward.sh

(3) End to end training

./scripts_train_val/train_e2e.sh




4. Validation

(1) inference of forward and backward networks

./scripts_train_val/val.sh

(2) inference of end-to-end network

./scripts_train_val/val_e2d.sh



