import os
import sys
from keras.utils.data_utils import get_file
ORIGIN = 'https://github.com/dluvizon/deephar/releases/download/'
def check_mpii_dataset():
version = 'v0.1'
try:
mpii_path = os.path.join(os.getcwd(), 'datasets/MPII/')
annot_path = get_file(mpii_path + 'annotations.mat',
ORIGIN + version + '/mpii_annotations.mat',
md5_hash='cc62b1bb855bf4866d19bc0637526930')
if os.path.isdir(mpii_path + 'images') is False:
raise Exception('MPII dataset (images) not found! '
'You must download it by yourself from '
'http://human-pose.mpi-inf.mpg.de')
except:
sys.stderr.write('Error checking MPII dataset!\n')
raise
def check_h36m_dataset():
version = 'v0.2'
try:
h36m_path = os.path.join(os.getcwd(), 'datasets/Human3.6M/')
annot_path = get_file(h36m_path + 'annotations.mat',
ORIGIN + version + '/h36m_annotations.mat',
md5_hash='4067d52db61737fbebdec850238d87dd')
if os.path.isdir(h36m_path + 'images') is False:
raise Exception('Human3.6M dataset (images) not found! '
'You must download it by yourself from '
'http://vision.imar.ro/human3.6m '
'and extract the video files!')
except:
sys.stderr.write('Error checking Human3.6M dataset!\n')
raise
def check_pennaction_dataset():
version = 'v0.3'
try:
penn_path = os.path.join(os.getcwd(), 'datasets/PennAction/')
annot_path = get_file(penn_path + 'annotations.mat',
ORIGIN + version + '/penn_annotations.mat',
md5_hash='b37a2e72c0ba308bd7ad476bc2aa4d33')
bbox_path = get_file(penn_path + 'penn_pred_bboxes_16f.json',
ORIGIN + version + '/penn_pred_bboxes_16f.json',
md5_hash='30b124a919185cb031b928bc6154fa9b')
if os.path.isdir(penn_path + 'frames') is False:
raise Exception('PennAction dataset (frames) not found! '
'You must download it by yourself from '
'http://dreamdragon.github.io/PennAction')
except:
sys.stderr.write('Error checking PennAction dataset!\n')
raise
def check_ntu_dataset():
try:
ntu_path = os.path.join(os.getcwd(), 'datasets/NTU/')
if os.path.isdir(ntu_path + 'images-small') is False:
raise Exception('NTU dataset (images-small) not found! '
'You must download it by yourself from '
'http://rose1.ntu.edu.sg/Datasets/actionRecognition.asp '
'and extract the video files. A helper Python script is '
'given for that in '
'datasets/NTU/extract-resize-videos.py')
if os.path.isdir(ntu_path + 'nturgb+d_numpy') is False:
raise Exception('NTU dataset (nturgb+d_numpy) not found! '
'Please download the annotations from '
'TODO [LINK] '
'and extract the file in datasets/NTU')
except:
sys.stderr.write('Error checking NTU dataset!\n')
raise
