Skip to content

Latest commit

 

History

History
98 lines (77 loc) · 3.38 KB

train.md

File metadata and controls

98 lines (77 loc) · 3.38 KB
Raw

Pre-reqs

Download required models

  1. (Same as demo) Download the pre-trained models. Place the models folder as a top-level directory.
wget http://angjookanazawa.com/cachedir/hmmr/hmmr_models.tar.gz && tar -xf hmmr_models.tar.gz

Download datasets.

Download these datasets:

If you use any of the data above, please cite the corresponding datasets and follow their licenses:

article{Loper:SIGASIA:2014,
  title = {{MoSh}: Motion and Shape Capture from Sparse Markers},
  author = {Loper, Matthew M. and Mahmood, Naureen and Black, Michael J.},
  journal = {ACM Transactions on Graphics, (Proc. SIGGRAPH Asia)},
  volume = {33},
  number = {6},
  pages = {220:1--220:13},
  publisher = {ACM},
  address = {New York, NY, USA},
  month = nov,
  year = {2014},
  url = {http://doi.acm.org/10.1145/2661229.2661273},
  month_numeric = {11}
}

TFRecord Generation

All the data has to be converted into TFRecords and saved to a DATA_DIR of your choice.

  1. Make DATA_DIR where you will save the tf_records. For ex:
mkdir ~/hmmr/tf_datasets/

  1. Edit prepare_datasets.sh, with paths to where you downloaded the datasets, and set DATA_DIR to the path to the directory you just made. Please read the comments in prepare_datasets.sh for each dataset. Some require running a pre-processing script, which is spelled out in prepare_datasets.sh

  2. Follow the direction in prepare_datasets.sh, for each dataset, uncomment each command and from the root HMMR directly (where README is), run prepare_datasets.sh:

sh prepare_datasets.sh

This takes a while! Start with UPenn and Human3.6M. Insta_variety and Vlog people are optional.

  1. In do_train.sh and/or src/config.py, set DATA_DIR to the path where you saved the tf_records.

Training

Now we can start training. A sample training script (with parameters used in the paper) is in do_train.sh.

Update the path to in the beginning of this script and run:

sh do_train.sh

The training write to a log directory that you can specify. Setup tensorboard to this directory to monitor the training progress like so: Teaser Image

It's important to visually monitor the training! Make sure that the images loaded look right.

config

do_train.sh will use the default weights/setting. But see src/config.py for more options.

Note on data.

Unfortunately, due to the licensing, the Human3.6M Mosh data is no longer available. This means you will not be able to reproduce exactly the public model since the supervision will not have the mosh values. Human3.6M data produced above will still have the 3D joint annotation, but not the SMPL values. We have re-trained our model on this setting where Human3.6M mosh data is not available, see doc/eval.md