Scripts for viewing Creatures Great and SMAL results.
-
Clone the repository with submodules and enter directory
git clone --recurse-submodules https://github.com/benjiebob/CreaturesResult cd CreaturesResult
-
Install the Neural Mesh Renderer submodule (which originated from my fork), which includes a render_points function
pip install -e neural_renderer
-
Download texture map (from smal/dog_texture.pkl) and a version of SMAL 2017 converted to NumPy (smal_CVPR2017_np.pkl) from my Google Drive and place under the smal folder
-
Download the preliminary Maggie result data
-
Install dependencies, particularly PyTorch (with cuda support)
-
Edit the IMAGE_DIR and RESULT_DIR paths in result_viewer.py to reflect your own file structure
-
Test the python3 script
python result_viewer.py
This work was completed in relation to the following paper:
@inproceedings{biggs2018creatures,
title={{C}reatures great and {SMAL}: {R}ecovering the shape and motion of animals from video},
author={Biggs, Benjamin and Roddick, Thomas and Fitzgibbon, Andrew and Cipolla, Roberto},
booktitle={ACCV},
year={2018}
}