This is the repository of "LSTM-KF: Long Short-Term Memory Kalman Filters: Recurrent Neural Estimators for Pose Regularization" presented at ICCV 2017, by Huseyin Coskun, Felix Achilles, Robert DiPietro, Nassir Navab, and Federico Tombari. You are free to use this code in non-commercial applications. For use in a publication or presentation, please cite the LSTM-KF.
tensorflow
python 2.7
numpy
h5py
NVIDIA CUDA 8.0 (for the GPU implementation)
If you have any questions about this, please reach out to me at: huseyin.coskun@tum.de
tensorflow NVIDIA CUDA 8.0 (for the GPU implementation)
User python train_h36m.py to train Human3.6m data set. You need Inception prediction that you can get via using python train.py We will provide pre-trined models for the InceptionV4. Feel free to participate to development of the code