DATA

• • There are 5 subjects for training and 2 for validation

• • There are a total of 15 actions performed by the actors

• • In the 'Train' ('Val') folder there are 2 sub-folders:

           ◦ 'IMG' ----> contains 35832 (19312) RGB images (JPG format) depicting single people in the Human3.6M laboratory setting %05d.jpg (from 00001.jpg to 35831.jpg/19312.jpg)

           ◦ 'POSE' ----> contains the corresponding 35832 (19312) pose (CSV format) information

  %05d.csv (from 00001.csv to 35831.csv/19312.csv)

  X00,Y00,Z00

  X01,Y01,Z01

  ...

  X15,Y15,Z15

  X16,Y16,Z16

  These are XYZ coordinates relative to the pelvis-joint (X00 = 0, Y00 = 0, Z00 = 0 is the pelvis-joint), expressed in millimeters (mm).

  • • The 17 joints - in order - are: {'Pelvis' 'RHip' 'RKnee' 'RAnkle' 'LHip' 'LKnee' 'LAnkle' 'Spine1' 'Neck' 'Head' 'Site' 'LShoulder' 'LElbow' 'LWrist' 'RShoulder' 'RElbow' 'RWrist};

VISUALIZATION

To visualize a pose associated with a particular image, run the following commands:

cd MY_PATH/Scripts

python plotSkel.py ../Train/POSE/14232.csv

VALIDATING

To validate the method on the validation data (in 'Val') follow these steps:

• 1. Assuming that the data is in 'MY_PATH', then for each image in Val/IMG, generate a .csv file with the same name, in a separate folder 'MY_PATH/Val/PRED'

• 2. EACH .csv file should have the same format as the train files:

%05d.csv (from 00001.csv to 19312.csv)

X00,Y00,Z00

X01,Y01,Z01

...

X15,Y15,Z15

X16,Y16,Z16

containing 17x3 predictions. Please NOTE that each skeleton is centered w.r.t to the pelvis: X00,Y00,Z00 should always be 0.

• 3. From the folder 'MY_PATH/Scripts', run the following command: "python validate.py -p MY_PATH/Val/PRED"

EXAMPLE RUN:

cd MY_PATH/Scripts

python validate.py -p ../Val/PRED

• 4. The error is computed as the mean MPJPE metric, expressed in mm.

  TESTING

  Download the test data first and follow the instructions from README_SUBMISSION