This folder contains script which can be used to parse the VIBE skeleton data of skeletics-152 and skeleton-mimetics
There is one main script:
- parse_vibe_data.py (Parses VIBE data into VaCNN and GCN format)
- VaCNN data is stored in .h5 format
- GCN data is stored in .npy format
GCN data is stored is most standard format used by most of the GCN based models.
The data is stored in N x C x T x V x M , where
- N = Number of samples
- C = Number of channels (C = 3 for 3d data)
- T = Number of Frames (T = 300 by default)
- V = Number of Joints (V = 25 by default)
- M = Number of Persons (M = 2 by default)
python3 parse_vibe_data.py INPUT_PATH OUTPUT_PATH
Where
- INPUT_PATH = path to the directory containing VIBE .json files
- OUTPUT_PATH = path where parsed data needs to be saved.
After the script is finished a vacnn_data.h5 file will be created with VaCNN data and labels.
Also two more files will be created, gcn_train_data.npy and gcn_train_label.npy which will contain data and label respectively.
(NOTE: To parse different splits of dataset give the input path accordingly)