Active Learning for Bayesian 3D Hand Pose Estimation

If you find this code useful please cite: https://arxiv.org/abs/2010.00694

Proposed pipeline

The code requires pre-processed hand detected and cropped images for the 3D Hand Pose Dataset: ICVL, NYU, BigHand2.2; This should be place in the "data/source/" directory The pre-processed dataset can be made available upon request([email protected]).

Before running, please check and tune the parameters in the config.py file and at the beginning of the main.py function.

Bayesian DeepPrior results with aleatoric and epistemic uncertainty

Qualitative analysis on the ICVL, NYU, BigHand2.2M datasets

https://www.youtube.com/playlist?list=PLYk-b7Fzd073O2ksuTSvxLNXiMEOMSaIh

Required Python3 libraries:

numpy
matplotlib
tensorflow
keras
opencv
skicit-learn
scipy
math
h5py
glob
gc
argparse