Skip to content

Repository for "Height and Weight Estimation From Unconstrained Images" paper.

Notifications You must be signed in to change notification settings

codesharpdev/DeepHeightWeight

 
 

Repository files navigation

Height and Weight Estimation From Unconstrained Images

Authors

Can Yilmaz Altinigne, Dorina Thanou and Radhakrishna Achanta

Abstract

We address the difficult problem of estimating the attributes of weight and height of individuals from pictures taken in completely unconstrained settings. We present a deep learning scheme that relies on simultaneous prediction of human silhouettes and skeletal joints as strong regularizers that improve the prediction of attributes such as height and weight. Apart from imparting robustness to the prediction of attributes, our regularization also allows for better visual interpretability of the attribute prediction. For height estimation, our method shows lower mean average error compared to the state of the art despite using a simpler approach. For weight estimation, which has hardly been addressed in the literature, we set a new benchmark. (Accepted to IEEE ICASSP 2020)

Image of Yaktocat

Reference

If you find this code useful in your research, please consider citing:

@inproceedings{altinigne2020height,
  title={Height and Weight Estimation from Unconstrained Images},
  author={Altinigne, Can Yilmaz and Thanou, Dorina and Achanta, Radhakrishna},
  booktitle={ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  pages={2298--2302},
  year={2020},
  organization={IEEE}
}

Trained Models in the Original Paper

Each file is nearly 2GB. Just download them and put them in /models folder.

Dependency

  • PyTorch = 1.2.0
  • Python = 3.6

Usage

In order to train height network, run train.py. For weight network run train_weight.py. Instructions are given as comment lines in the header. You can find the performance of models on height and weight estimation test sets in Results_HW.ipynb notebook.

Environment

You can use environment.yml file to create a Conda environment to run experiments. You can create a new environment using this command. conda env create -f environment.yml -p [PATH]

Weight Dataset

We will soon share the weight dataset that we created and used in our experiments.

About

Repository for "Height and Weight Estimation From Unconstrained Images" paper.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 83.6%
  • Python 16.4%