Awesome

Three-D Safari: Learning to Estimate Zebra Pose, Shape, and Texture from Images "In the Wild"

Silvia Zuffi¹, Angjoo Kanazawa², Tanya Berger-Wolf³, Michael J. Black⁴

¹IMATI-CNR, Milan, Italy, ²University of California, Berkeley, ³University of Illinois at Chicago, ⁴Max Planck Institute for Intelligent Systems, Tuebingen, Germany

In ICCV 2019

paper

suppmat

Requirements

• Python 2.7
• PyTorch tested on version 0.5.0

Installation

Note that the following warning has been issued: "Pillow before 8.1.1 allows attackers to cause a denial of service (memory consumption) because the reported size of a contained image is not properly checked for an ICO container, and thus an attempted memory allocation can be very large."

Setup virtualenv

virtualenv venv_smalst
source venv_smalst/bin/activate
pip install -U pip
deactivate
source venv_smalst/bin/activate
pip install -r requirements.txt

Install Neural Mesh Renderer and Perceptual loss

cd external;
bash install_external.sh

Install SMPL model

download the SMPL model and create a directory smpl_webuser under the smalst/smal_model directory

Download data

• Trained network

• Training data

• Test data

• Validation data

The test and validation data are images collected in The Great Grevy's Rally 2018

Place the downloaded network pred_net_186.pth in the folder cachedir/snapshots/smal_net_600/

Usage

See the script in smalst/script directory for training and testing

Notes

The code in this repository is widely based on the project https://github.com/akanazawa/cmr

Citation

If you use this code please cite

@inproceedings{Zuffi:ICCV:2019,
  title = {Three-D Safari: Learning to Estimate Zebra Pose, Shape, and Texture from Images "In the Wild"},
  author = {Zuffi, Silvia and Kanazawa, Angjoo and Berger-Wolf, Tanya and Black, Michael J.},
  booktitle = {International Conference on Computer Vision},
  month = oct,
  year = {2019},
  month_numeric = {10}
}