Data

MPII Dataset

Official link provides data in matlab format.

Annotation data in pythonic format processed by @pulkitag can be downloaded from this link. The folder contains two files:

• train_data.pkl
• val_data.pkl

that reflect the training and validation data used in the IEF experiments.

Annotations can be extracted in the following manner:

import pickle
data = pickle.load(open('train_data.pkl', 'r'))
#Name of the image (images can be downloaded from official MPII link above)
imName  = data['imName'] 
#The torso point on the body of the person
bodyPos = data['bodyPos']
#len(bodyPos) - number of people
#bodyPos[i]   - the torso point of the ith person
#Scale of the human
bodyScale = data['bodyScale']
#Keypoints annotation
kpts      = data['kpts']