def getPCL(self, dpt, T):
"""
Get pointcloud from frame
:param dpt: depth image
:param T: 2D transformation of crop
"""
return Blender2Importer.depthToPCL(dpt, T)
def __init__(self, imgSeqs=None, basepath=None, localCache=True):
"""
constructor
"""
super(Blender2Dataset, self).__init__(imgSeqs, localCache)
if basepath is None:
basepath = '../../data/Blender/'
self.lmi = Blender2Importer(basepath)
if start_idx is None:
print(
'Start frame index can be specified: -s <start_idx> or enter now:')
start_idx = input().lower()
if len(start_idx.strip()) == 0:
start_idx = 0
else:
start_idx = int(start_idx)
else:
print('Start frame index is {}'.format(start_idx))
rng = numpy.random.RandomState(23455)
if person == 'hpseq_loop_mv':
di = Blender2Importer('../data/Blender/', useCache=False)
Seq2 = di.loadSequence(person, camera=0, shuffle=False)
hpe = Blender2HandposeEvaluation([j.gt3Dorig for j in Seq2.data],
[j.gt3Dorig for j in Seq2.data])
for idx, seq in enumerate(Seq2.data):
Seq2.data[idx] = seq._replace(com=numpy.zeros((3, )))
else:
raise NotImplementedError("")
# we need to detect all files
subset_idxs = []
output_path = di.basepath
filename_dets = output_path + person + '/detections_' + username + '.txt'
filename_log = output_path + person + '/detecttool_log_' + username + '.txt'
from data.transformations import transformPoint2D
from data.importers import Blender2Importer
from data.dataset import Blender2Dataset
from util.handpose_evaluation import Blender2HandposeEvaluation
if __name__ == '__main__':
eval_prefix = 'BLEN_SA'
if not os.path.exists('./eval/' + eval_prefix + '/'):
os.makedirs('./eval/' + eval_prefix + '/')
rng = numpy.random.RandomState(23455)
print("create data")
di = Blender2Importer('../data/Blender/')
Seq1_0 = di.loadSequence('hpseq_loop_mv', camera=0, shuffle=False)
trainSeqs = [Seq1_0]
# create training data
trainDataSet = Blender2Dataset(trainSeqs)
dat = []
gt = []
for seq in trainSeqs:
d, g = trainDataSet.imgStackDepthOnly(seq.name)
dat.append(d)
gt.append(g)
train_data = numpy.concatenate(dat)
train_gt3D = numpy.concatenate(gt)
mb = (train_data.nbytes) / (1024 * 1024)