Skip to content

Latest commit

 

History

History

cafi_net

README.md

Training and Testing Cafinet

This README contains instructions to train and test the Cafinent.

Loading the environment

Make sure you have Anaconda or Miniconda installed before you proceed to load this environment.

# Creating the conda environment and loading it
conda env create -f environment.yml
conda activate Cafinet_torch

Training and testing .

Dataset

# Download dataset
mkdir data
cd data
wget https://nerf-fields.s3.amazonaws.com/final_fields/final_res_32.zip
# Unzip dataset
unzip final_res_32.zip

Training

  1. In configs/Canonical_fields.yaml change the dataset path to the downloaded dataset.
# In configs/Canonical_fields.yaml
dataset:
  dataset_path: <change path to to training dataset>
val_dataset:
  dataset_path:: <change path to to validation dataset>
  1. Run the code below
# Run the code to train
CUDA_VISIBLE_DEVICES=0 python main.py

Testing

  1. The test script tests the model on the validation set and saves the output as ply files for visualization.
# Test the trained model
# weight files are stored at path outputs/<date_of_run_train>/<time_of_run_train>/checkpoints/ 
CUDA_VISIBLE_DEVICES=0 python3 tester.py 'test.weights="<model_weights_path>"' 'test.skip=1'
  1. After running the test script you will find a new directory with stored pointclouds at location outputs/<date_of_run_test>/<time_of_run_test>/pointclouds/
  2. To visualize the pointcliuds use the below script
python vis_utis.py --base_path <path containg the pointclouds> --pcd <*pattern for the point coluds>