Download

MPII Human Pose Dataset, Version 1.0
Copyright 2015 Max Planck Institute for Informatics
Licensed under the Simplified BSD License [see bsd.txt]
 

We are making the annotations and the corresponding code freely available for research purposes. Commercial use is not allowed due to the fact that the authors do not have the copyright for the images themselves.

You can download the images and annotations from the MPII Human Pose benchmark here:

 Images (12.9 GB)

 Annotations (12.5 MB)

Below you can download short videos including preceding and following frames for each image. The videos are split for download into 25 batches ~17 GB each:

 Batch 1   Batch 2   Batch 3   Batch 4   Batch 5 

 Batch 6   Batch 7   Batch 8   Batch 9   Batch 10 

 Batch 11   Batch 12   Batch 13   Batch 14   Batch 15 

 Batch 16   Batch 17   Batch 18   Batch 19   Batch 20 

 Batch 21   Batch 22   Batch 23   Batch 24   Batch 25

 Image - video mapping (239 KB)

Annotation description

Annotations are stored in a matlab structure RELEASE having following fields

  • .annolist(imgidx) - annotations for image imgidx

    • .image.name - image filename
    • .annorect(ridx) - body annotations for a person ridx
      • .x1, .y1, .x2, .y2 - coordinates of the head rectangle
      • .scale - person scale w.r.t. 200 px height
      • .objpos - rough human position in the image
      • .annopoints.point - person-centric body joint annotations
        • .x, .y - coordinates of a joint
        • id - joint id (0 - r ankle, 1 - r knee, 2 - r hip, 3 - l hip, 4 - l knee, 5 - l ankle, 6 - pelvis, 7 - thorax, 8 - upper neck, 9 - head top, 10 - r wrist, 11 - r elbow, 12 - r shoulder, 13 - l shoulder, 14 - l elbow, 15 - l wrist)
        • is_visible - joint visibility
    • .vidx - video index in video_list
    • .frame_sec - image position in video, in seconds
  • img_train(imgidx) - training/testing image assignment

  • single_person(imgidx) - contains rectangle id ridx of sufficiently separated individuals

  • act(imgidx) - activity/category label for image imgidx

    • act_name - activity name
    • cat_name - category name
    • act_id - activity id
  • video_list(videoidx) - specifies video id as is provided by YouTube. To watch video on youtube go to https://www.youtube.com/watch?v=video_list(videoidx)