def get_dataflow(path, is_train):
ds = SynthHands(path, is_train) # read data from lmdb
if is_train:
ds = MapData(ds, read_image_url)
ds = MapData(ds, pose_to_img)
ds = PrefetchData(ds, 1000, multiprocessing.cpu_count() * 1)
else:
ds = MultiThreadMapData(ds, num_thread=16, map_func=read_image_url, buffer_size=1000)
ds = MapData(ds, pose_to_img)
ds = PrefetchData(ds, 100, multiprocessing.cpu_count() // 4)
return ds
last_gs_num2 = gs_num
sample_image = [enqueuer.last_dp[0][i] for i in range(4)]
outputMat = sess.run(outputs,
feed_dict={
q_inp:
np.array((sample_image + val_image) *
max(1,
(args.batchsize // 8)))
})
heatMat = outputMat[:, :, :, :19]
sample_results = []
for i in range(len(sample_image)):
test_result = SynthHands.display_image(sample_image[i],
heatMat[i],
as_numpy=True)
test_result = cv2.resize(test_result, (640, 640))
test_result = test_result.reshape([640, 640,
3]).astype(float)
sample_results.append(test_result)
test_results = []
for i in range(len(val_image)):
test_result = SynthHands.display_image(
val_image[i],
heatMat[len(sample_image) + i],
as_numpy=True)
test_result = cv2.resize(test_result, (640, 640))
test_result = test_result.reshape([640, 640,
3]).astype(float)
def _get_dataflow_onlyread(path, is_train):
ds = SynthHands(path, is_train) # read data from lmdb
ds = MapData(ds, read_image_url)
ds = MapData(ds, pose_to_img)
# ds = PrefetchData(ds, 1000, multiprocessing.cpu_count() * 4)
return ds
return [
meta_l[0].img.astype(np.float16),
meta_l[0].get_heatmap(target_size=(_network_w // _scale, _network_h // _scale))
]
if __name__ == '__main__':
os.environ['CUDA_VISIBLE_DEVICES'] = ''
# df = get_dataflow('/data/public/rw/coco/annotations', True, '/data/public/rw/coco/')
df = _get_dataflow_onlyread('/home/marcelo/hands/hand_labels_synth', True)
# df = get_dataflow('/root/coco/annotations', False, img_path='http://gpu-twg.kakaocdn.net/braincloud/COCO/')
from tensorpack.dataflow.common import TestDataSpeed
TestDataSpeed(df).start()
with tf.Session() as sess:
df.reset_state()
t1 = time.time()
for idx, dp in enumerate(df.get_data()):
if idx == 0:
for d in dp:
logger.info('%d dp shape={}'.format(d.shape))
print(time.time() - t1)
t1 = time.time()
SynthHands.display_image(dp[0], dp[1].astype(np.float32))
pass
logger.info('done')
print(
"Belief maps for last stage, one for each keypoint plus and additional one for the background"
)
fig, ax = plt.subplots(nrows=5, ncols=5)
index = 0
for row in ax:
for col in row:
col.imshow(last_stage[:, :, index])
index += 1
if index >= 22:
break
plt.show()
print("Draws last stage belief maps on top of image")
test_result = SynthHands.display_image(img, last_stage, as_numpy=True)
test_result = cv2.cvtColor(test_result, cv2.COLOR_RGB2BGR)
cv2.imshow("Belief Maps", test_result)
cv2.waitKey(0)
print("Draws each stage belief maps")
fig = plt.figure()
for i, stage_out in enumerate(stages_outs):
stage_out = stage_out[0]
a = fig.add_subplot(3, 2, i + 1)
a.set_title('Stage #{}'.format(i + 1))
plt.imshow(SynthHands.get_bgimg(img,
target_size=(stage_out.shape[1],
stage_out.shape[0])),
alpha=0.5)
tmp = np.amax(stage_out, axis=2)