def start_videoanalyser():
print("Starting DeepLabCut")
model = load_model(
r"D:\DeepPoseKit-Data-master\datasets\fly\best_model_densenet.h5")
experiment_enabled = False
video_output = True
if experiment_enabled:
print("Initializing experiment")
experiment = ExampleExperiment()
experiment.start_experiment()
# some variables initialization
all_rows = []
index = 0
while video.isOpened():
ret, frame = video.read()
if ret:
scmap, locref, pose = get_pose(frame, config, sess, inputs,
outputs)
peaks = find_local_peaks_new(scmap, locref, ANIMALS_NUMBER, config)
skeletons = calculate_skeletons(peaks, ANIMALS_NUMBER)
if skeletons:
for skeleton in skeletons:
if experiment_enabled:
result, response = experiment.check_skeleton(
frame, skeleton)
plot_triggers_response(frame, response)
out_frame = plot_bodyparts(frame, skeletons)
else:
out_frame = frame
cv2.imshow('stream', out_frame)
if video_output:
video_file.write(out_frame)
if experiment_enabled:
all_rows.append(
create_row(index, skeletons, experiment_enabled,
experiment.get_trial()))
else:
all_rows.append(
create_row(index, skeletons, experiment_enabled, None))
index += 1
else:
break
if cv2.waitKey(1) & 0xFF == ord('q'):
break
if experiment_enabled:
experiment.stop_experiment()
if video_output:
print('Saving analyzed video for {}'.format(video_name))
video_file.release()
video.release()
create_dataframes(all_rows)
def get_pose_mp(input_q, output_q):
"""
Process to be used for each camera/DLC stream of analysis
Designed to be run in an infinite loop
:param input_q: index and corresponding frame
:param output_q: index and corresponding analysis
"""
if MODEL_ORIGIN in ('DLC', 'MADLC'):
config, sess, inputs, outputs = load_deeplabcut()
while True:
if input_q.full():
index, frame = input_q.get()
start_time = time.time()
if MODEL_ORIGIN == 'DLC':
scmap, locref, pose = get_pose(frame, config, sess, inputs, outputs)
peaks = find_local_peaks_new(scmap, locref, ANIMALS_NUMBER, config)
#Use the line below to use raw DLC output rather then DLStream optimization
# peaks = pose
if MODEL_ORIGIN == 'MADLC':
peaks = get_ma_pose(frame, config, sess, inputs, outputs)
analysis_time = time.time() - start_time
output_q.put((index, peaks, analysis_time))
elif MODEL_ORIGIN == 'DLC-LIVE':
dlc_live = load_dlc_live()
while True:
if input_q.full():
index, frame = input_q.get()
start_time = time.time()
if not dlc_live.is_initialized:
peaks = dlc_live.init_inference(frame)
else:
peaks = dlc_live.get_pose(frame)
analysis_time = time.time() - start_time
output_q.put((index, peaks, analysis_time))
elif MODEL_ORIGIN == 'DEEPPOSEKIT':
predict_model = load_dpk()
while True:
if input_q.full():
index, frame = input_q.get()
start_time = time.time()
frame = frame[..., 1][..., None]
st_frame = np.stack([frame])
prediction = predict_model.predict(st_frame, batch_size=1, verbose=True)
peaks = prediction[0, :, :2]
analysis_time = time.time() - start_time
output_q.put((index,peaks,analysis_time))
elif MODEL_ORIGIN == 'SLEAP':
sleap_model = load_sleap()
while True:
if input_q.full():
index, frame = input_q.get()
start_time = time.time()
input_frame = frame[:, :, ::-1]
#this is weird, but without it, it does not seem to work...
frames = np.array([input_frame])
prediction = sleap_model.predict(frames[[0]], batch_size=1)
#check if this is multiple animal instances or single animal model
if sleap_model.name == 'single_instance_inference_model':
#get predictions (wrap it again, so the behavior is the same for both model types)
peaks = np.array([prediction['peaks'][0, :]])
else:
peaks = prediction['instance_peaks'][0, :]
analysis_time = time.time() - start_time
output_q.put((index,peaks,analysis_time))
else:
raise ValueError(f'Model origin {MODEL_ORIGIN} not available.')
def get_pose_mp(input_q, output_q):
"""
Process to be used for each camera/DLC stream of analysis
Designed to be run in an infinite loop
:param input_q: index and corresponding frame
:param output_q: index and corresponding analysis
"""
if MODEL_ORIGIN in ("DLC", "MADLC"):
config, sess, inputs, outputs = load_deeplabcut()
while True:
if input_q.full():
index, frame = input_q.get()
start_time = time.time()
if MODEL_ORIGIN == "DLC":
scmap, locref, pose = get_pose(frame, config, sess,
inputs, outputs)
if USE_DLSTREAM_POSTURE_DETECTION:
""" This is a legacy function that was used in earlier versions"""
peaks = find_local_peaks_new(
scmap, locref, ANIMALS_NUMBER, config)
# Use the line below to use raw DLC output rather then DLStream optimization
else:
peaks = pose
if MODEL_ORIGIN == "MADLC":
peaks = get_ma_pose(frame, config, sess, inputs,
outputs)
analysis_time = time.time() - start_time
output_q.put((index, peaks, analysis_time))
elif MODEL_ORIGIN == "DLC-LIVE":
dlc_live = load_dlc_live()
while True:
if input_q.full():
index, frame = input_q.get()
start_time = time.time()
if not dlc_live.is_initialized:
peaks = dlc_live.init_inference(frame)
else:
peaks = dlc_live.get_pose(frame)
analysis_time = time.time() - start_time
output_q.put((index, peaks, analysis_time))
elif MODEL_ORIGIN == "DEEPPOSEKIT":
predict_model = load_dpk()
while True:
if input_q.full():
index, frame = input_q.get()
start_time = time.time()
frame = frame[..., 1][..., None]
st_frame = np.stack([frame])
prediction = predict_model.predict(st_frame,
batch_size=1,
verbose=True)
peaks = prediction[0, :, :2]
analysis_time = time.time() - start_time
output_q.put((index, peaks, analysis_time))
elif MODEL_ORIGIN == "SLEAP":
sleap_model = load_sleap()
while True:
if input_q.full():
index, frame = input_q.get()
start_time = time.time()
# Make sure image is (1, height, width, channels) and uint8
# (height, width) -> (height, width, 1)
frame = np.expand_dims(
frame, axis=-1) if frame.ndim == 2 else frame
# (height, width, channels) -> (1, height, width, channels)
frame = np.expand_dims(
frame, axis=0) if frame.ndim == 3 else frame
# predict_on_batch is MUCH faster as it does not retrace the model graph for same size inputs
pred = sleap_model.predict_on_batch(frame)
try:
peaks = pred["instance_peaks"][
0] # (n_poses, n_nodes, 2)
except KeyError:
# necessary for old sleap versions where single_instance models have different key naming
peaks = pred["peaks"]
analysis_time = time.time() - start_time
output_q.put((index, peaks, analysis_time))
else:
raise ValueError(f"Model origin {MODEL_ORIGIN} not available.")