Human3.6M - Dataset processing and visualization

A library for processing and generating readable annotations for Human3.6M dataset.

Steps

1. Extract all dataset files

|-compressed
|----annotations.zip
|----s_01_act_02_subact_01_ca_01.zip
|----s_01_act_02_subact_01_ca_02.zip
|----s_01_act_02_subact_01_ca_03.zip
|----s_01_act_02_subact_01_ca_04.zip
|----...

python extract_all.py

First put all dataset ZIP files in to directory "compressed". The files will be extracted to folder "extracted"

2. Generate annotations

python extract_annotations.py

subject1_annotations.json
subject5_annotations.json
subject6_annotations.json
subject7_annotations.json
subject8_annotations.json
subject9_annotations.json
subject10_annotations.json
subject11_annotations.json

The annotations will be generated for each subject as json file in "extracted" folder.

To download GENERATED JSON files please email at [email protected].

3D Pose Visualization

To visualize a pose in 3D

python visualize_3d.py

2D Pose Visualization

To visualize a pose in 2D

python visualize_2d.py