def test(args):
# augmentation setting
trans = Transform(padding=[args.crop_pad_inf, args.crop_pad_sup],
flip=args.flip,
size=args.size,
shift=args.shift,
lcn=args.lcn)
# test data
test_fn = '%s/test_joints.csv' % args.datadir
test_dl = np.array([l.strip() for l in open(test_fn).readlines()])
# load model
if args.gpu >= 0:
cuda.init(args.gpu)
model = load_model(args)
if args.gpu >= 0:
model.to_gpu()
else:
model.to_cpu()
# create output dir
epoch = int(re.search(ur'epoch_([0-9]+)', args.param).groups()[0])
result_dir = os.path.dirname(args.param)
out_dir = '%s/test_%d' % (result_dir, epoch)
if not os.path.exists(out_dir):
os.makedirs(out_dir)
out_log = '%s.log' % out_dir
fp = open(out_log, 'w')
mean_error = 0.0
N = len(test_dl)
for i in range(0, N, args.batchsize):
lines = test_dl[i:i + args.batchsize]
input_data, labels = load_data(trans, args, lines)
if args.gpu >= 0:
input_data = cuda.to_gpu(input_data.astype(np.float32))
labels = cuda.to_gpu(labels.astype(np.float32))
_, preds = model.forward(input_data, labels, train=False)
if args.gpu >= 0:
preds = cuda.to_cpu(preds.data)
input_data = cuda.to_cpu(input_data)
labels = cuda.to_cpu(labels)
for n, line in enumerate(lines):
img_fn = line.split(',')[args.fname_index]
img = input_data[n].transpose((1, 2, 0))
pred = preds[n]
img_pred, pred = trans.revert(img, pred)
# turn label data into image coordinates
label = labels[n]
img_label, label = trans.revert(img, label)
# calc mean_error
error = np.linalg.norm(pred - label) / len(pred)
mean_error += error
# create pred, label tuples
img_pred = np.array(img_pred.copy())
img_label = np.array(img_label.copy())
pred = [tuple(p) for p in pred]
label = [tuple(p) for p in label]
# all limbs
img_label = draw_joints(img_label, label, args.draw_limb,
args.text_scale)
img_pred = draw_joints(img_pred, pred, args.draw_limb,
args.text_scale)
msg = '{:5}/{:5} {}\terror:{}\tmean_error:{}'.format(
i + n, N, img_fn, error, mean_error / (i + n + 1))
print(msg, file=fp)
print(msg)
fn, ext = os.path.splitext(img_fn)
tr_fn = '%s/%d-%d_%s_pred%s' % (out_dir, i, n, fn, ext)
la_fn = '%s/%d-%d_%s_label%s' % (out_dir, i, n, fn, ext)
cv.imwrite(tr_fn, img_pred)
cv.imwrite(la_fn, img_label)
parser.add_argument('--lcn', type=bool, default=True,
help='local contrast normalization for data'
'augmentation')
parser.add_argument('--joint_num', type=int, default=7)
parser.add_argument('--fname_index', type=int, default=0,
help='the index of image file name in a csv line')
args = parser.parse_args()
print(args)
# augmentation setting
trans = Transform(args)
# test data
#test_fn = '%s/test_joints.csv' % args.data_dir
#test_dl = np.array([l.strip() for l in open(test_fn).readlines()])
result_dir = '%s/test_trans/' % args.data_dir
if not os.path.exists(result_dir):
os.makedirs(result_dir)
#for i, line in enumerate(test_dl):
for i, line in enumerate(glob.glob('%s/images/*' % args.data_dir)):
orig, input_data, label = load_data(trans, args, line)
input_data = input_data.transpose((0, 2, 3, 1))[0].astype(np.float32)
label = label.astype(np.float32).flatten()
cv.imwrite('%s/%d_orig.jpg' % (result_dir, i), orig)
img, label = trans.revert(input_data, label)
label = [tuple(l) for l in label]
pose = draw_joints(input_data.copy(), label)
pose = np.array(pose.copy())
cv.imwrite('%s/%d_pose.jpg' % (result_dir, i), pose)
def video_test(args):
# augmentation setting
trans = Transform(args)
# test data
test_fn = args.datadir
#test_dl = np.array([l.strip() for l in open(test_fn).readlines()])
# load model
if args.gpu >= 0:
cuda.get_device(args.gpu).use()
model = load_model(args)
if args.gpu >= 0:
model.to_gpu()
else:
model.to_cpu()
# create output dir
epoch = int(re.search('epoch-([0-9]+)', args.param).groups()[0])
result_dir = os.path.dirname(args.param)
out_dir = '%s/test_%d' % (result_dir, epoch)
if not os.path.exists(out_dir):
os.makedirs(out_dir)
out_log = '%s.log' % out_dir
fp = open(out_log, 'w')
input_data_all, labels_all = load_video(trans, args)
preds_all = []
for i in range(0, len(input_data_all), args.batchsize):
if args.gpu >= 0:
input_data = cuda.to_gpu(
input_data_all[i:i + args.batchsize].astype(np.float32))
labels = cuda.to_gpu(labels_all[i:i + args.batchsize].astype(
np.float32))
x = Variable(input_data, volatile=True)
t = Variable(labels, volatile=True)
model(x, t)
if args.gpu >= 0:
preds = cuda.to_cpu(model.pred.data)
input_data = cuda.to_cpu(input_data)
labels = cuda.to_cpu(labels)
for n in xrange(len(input_data)):
img = input_data[n].transpose((1, 2, 0))
pred = preds[n]
img_pred, pred = trans.revert(img, pred)
# turn label data into image coordinates
label = labels[n]
img_label, label = trans.revert(img, label)
# create pred, label tuples
img_pred = np.array(img_pred.copy())
pred = [tuple(p) for p in pred]
# all limbs
img_pred = draw_joints(img_pred, pred, args.draw_limb,
args.text_scale)
tr_fn = 'tmp/%d.png' % (i + n + 1)
cv.imwrite(tr_fn, img_pred)
preds_all.append(pred)
save_vid_res(preds_all, out_dir, args.vidfile.split(".")[0])
def test(args):
# augmentation setting
trans = Transform(args)
# test data
test_fn = '%s/test_joints.csv' % args.datadir
test_dl = np.array([l.strip() for l in open(test_fn).readlines()])
# load model
if args.gpu >= 0:
cuda.get_device(args.gpu).use()
model = load_model(args)
if args.gpu >= 0:
model.to_gpu()
else:
model.to_cpu()
# create output dir
epoch = int(re.search('epoch-([0-9]+)', args.param).groups()[0])
result_dir = os.path.dirname(args.param)
out_dir = '%s/test_%d' % (result_dir, epoch)
if not os.path.exists(out_dir):
os.makedirs(out_dir)
out_log = '%s.log' % out_dir
fp = open(out_log, 'w')
mean_error = np.zeros(7)
N = len(test_dl)
for i in range(0, N, args.batchsize):
lines = test_dl[i:i + args.batchsize]
input_data, labels = load_data(trans, args, lines)
if args.gpu >= 0:
input_data = cuda.to_gpu(input_data.astype(np.float32))
labels = cuda.to_gpu(labels.astype(np.float32))
x = Variable(input_data, volatile=True)
t = Variable(labels, volatile=True)
model(x, t)
if args.gpu >= 0:
preds = cuda.to_cpu(model.pred.data)
input_data = cuda.to_cpu(input_data)
labels = cuda.to_cpu(labels)
for n, line in enumerate(lines):
img_fn = line.split(',')[args.fname_index]
img = input_data[n].transpose((1, 2, 0))
pred = preds[n]
img_pred, pred = trans.revert(img, pred)
# turn label data into image coordinates
label = labels[n]
img_label, label = trans.revert(img, label)
# calc mean_error
error = np.linalg.norm(pred - label, axis=1)
mean_error += error
# create pred, label tuples
img_pred = np.array(img_pred.copy())
img_label = np.array(img_label.copy())
pred = [tuple(p) for p in pred]
label = [tuple(p) for p in label]
# all limbs
img_label = draw_joints(img_label, label, args.draw_limb,
args.text_scale)
img_pred = draw_joints(img_pred, pred, args.draw_limb,
args.text_scale)
msg = '{:5}/{:5} {}\tshoulderl:{}\tshoulderr:{}\tnose:{}\telbowl:{}\telbowr:{}\twristl:{}\twristr:{}'.format(
i + n, N, img_fn, error[2], error[4], error[3], error[1],
error[5], error[0], error[6])
print(msg, file=fp)
print(msg)
fn, ext = os.path.splitext(img_fn.split('/')[-1])
tr_fn = '%s/%d-%d_%s_pred%s' % (out_dir, i, n, fn, ext)
la_fn = '%s/%d-%d_%s_label%s' % (out_dir, i, n, fn, ext)
cv.imwrite(tr_fn, img_pred)
cv.imwrite(la_fn, img_label)
parser.add_argument('--lcn', type=bool, default=True,
help='local contrast normalization for data augmentation')
parser.add_argument('--joint_num', type=int, default=7)
args = parser.parse_args()
print(args)
# augmentation setting
trans = Transform(padding=[args.crop_pad_inf, args.crop_pad_sup],
flip=args.flip,
size=args.size,
shift=args.shift,
lcn=args.lcn)
# test data
test_fn = '%s/test_joints.csv' % args.data_dir
test_dl = np.array([l.strip() for l in open(test_fn).readlines()])
result_dir = '%s/test_trans' % args.data_dir
if not os.path.exists(result_dir):
os.makedirs(result_dir)
for i, line in enumerate(test_dl):
orig, input_data, label = load_data(trans, args, line)
input_data = input_data.transpose((0, 2, 3, 1))[0].astype(np.float32)
label = label.astype(np.float32).flatten()
cv.imwrite('%s/%d_orig.jpg' % (result_dir, i), orig)
img, label = trans.revert(input_data, label)
label = [tuple(l) for l in label]
pose = draw_joints(input_data.copy(), label)
pose = np.array(pose.copy())
cv.imwrite('%s/%d_pose.jpg' % (result_dir, i), pose)
def test(args):
# test data
test_fn = '%s/test_joints.csv' % args.datadir
test_dl = np.array([l.strip() for l in open(test_fn).readlines()])
# load model
if args.gpu >= 0:
cuda.get_device(args.gpu).use()
model = load_model(args)
if args.gpu >= 0:
model.to_gpu(args.gpu)
# create output dir
epoch = int(re.search('epoch-([0-9]+)', args.param).groups()[0])
result_dir = os.path.dirname(args.param)
out_dir = '%s/test_%d' % (result_dir, epoch)
if not os.path.exists(out_dir):
os.makedirs(out_dir)
out_log = '%s.log' % out_dir
fp = open(out_log, 'w')
mean_error = 0.0
N = len(test_dl)
for i in range(0, N, args.batchsize):
lines = test_dl[i:i + args.batchsize]
input_data, labels = load_data(trans, args, lines)
if args.gpu >= 0:
input_data = cuda.to_gpu(input_data.astype(np.float32))
labels = cuda.to_gpu(labels.astype(np.float32))
else:
input_data = input_data.astype(np.float32)
labels = labels.astype(np.float32)
x = Variable(input_data, volatile=True)
t = Variable(labels, volatile=True)
model(x, t)
if args.gpu >= 0:
preds = cuda.to_cpu(model.pred.data)
input_data = cuda.to_cpu(input_data)
labels = cuda.to_cpu(labels)
else:
preds = model.pred.data
for n, line in enumerate(lines):
img_fn = line.split(',')[args.fname_index]
img = input_data[n].transpose((1, 2, 0))
pred = preds[n]
img_pred, pred = trans.revert(img, pred)
# turn label data into image coordinates
label = labels[n]
img_label, label = trans.revert(img, label)
# calc mean_error
error = np.linalg.norm(pred - label) / len(pred)
mean_error += error
# create pred, label tuples
img_pred = np.array(img_pred.copy())
img_label = np.array(img_label.copy())
pred = [tuple(p) for p in pred]
label = [tuple(p) for p in label]
# all limbs
img_label = draw_joints(
img_label, label, args.draw_limb, args.text_scale)
img_pred = draw_joints(
img_pred, pred, args.draw_limb, args.text_scale)
msg = '{:5}/{:5} {}\terror:{}\tmean_error:{}'.format(
i + n, N, img_fn, error, mean_error / (i + n + 1))
print(msg, file=fp)
print(msg)
fn, ext = os.path.splitext(img_fn)
tr_fn = '%s/%d-%d_%s_pred%s' % (out_dir, i, n, fn, ext)
la_fn = '%s/%d-%d_%s_label%s' % (out_dir, i, n, fn, ext)
cv.imwrite(tr_fn, img_pred)
cv.imwrite(la_fn, img_label)
def demo(args):
# augmentation setting
trans = Transform(args)
# test data
test_dl = sorted(glob.glob('%s/*' % args.datadir))
# load model
if args.gpu >= 0:
cuda.get_device(args.gpu).use()
model = load_model(args)
if args.gpu >= 0:
model.to_gpu()
else:
model.to_cpu()
# create output dir
epoch = int(re.search('epoch-([0-9]+)', args.param).groups()[0])
out_dir = '%s/' % (args.resultdir)
if not os.path.exists(out_dir):
os.makedirs(out_dir)
out_log = '%s.log' % out_dir
fp = open(out_log, 'w')
N = len(test_dl)
with open('%s/joint_coords.csv' % out_dir, 'wb') as csvfile:
writer = csv.writer(csvfile)
for i in range(0, N, args.batchsize):
lines = test_dl[i:i + args.batchsize]
input_data, labels = load_data_demo(trans, args, lines)
if args.gpu >= 0:
input_data = cuda.to_gpu(input_data.astype(np.float32))
labels = cuda.to_gpu(labels.astype(np.float32))
x = Variable(input_data, volatile=True)
t = Variable(labels, volatile=True)
model(x, t)
if args.gpu >= 0:
preds = cuda.to_cpu(model.pred.data)
input_data = cuda.to_cpu(input_data)
labels = cuda.to_cpu(labels)
for n, line in enumerate(lines):
img_fn = line.split(',')[args.fname_index]
img = input_data[n].transpose((1, 2, 0))
pred = preds[n]
img_pred, pred = trans.revert(img, pred)
# create pred
img_pred = np.array(img_pred.copy())
pred = [tuple(p) for p in pred]
writer.writerow(pred)
# all limbs
img_pred = draw_joints(img_pred, pred, args.draw_limb,
args.text_scale)
fn, ext = os.path.splitext(img_fn.split('/')[-1])
tr_fn = '%s/%s_pred%s' % (out_dir, fn, ext)
cv.imwrite(tr_fn, img_pred)
