import numpy as np
import time
import cv2
import socket
import struct
import json
from input_reader import InputReader, VideoReader, DShowCaptureReader, try_int
from tracker import Tracker, get_model_base_path
if args.benchmark > 0:
model_base_path = get_model_base_path(args.model_dir)
im = cv2.imread(os.path.join(model_base_path, "benchmark.bin"), cv2.IMREAD_COLOR)
results = []
for model_type in [3, 2, 1, 0, -1, -2, -3]:
tracker = Tracker(224, 224, threshold=0.1, max_threads=args.max_threads, max_faces=1, discard_after=0, scan_every=0, silent=True, model_type=model_type, model_dir=args.model_dir, no_gaze=(model_type == -1), detection_threshold=0.1, use_retinaface=0, max_feature_updates=900, static_model=True if args.no_3d_adapt == 1 else False)
tracker.detected = 1
tracker.faces = [(0, 0, 224, 224)]
total = 0.0
for i in range(100):
start = time.perf_counter()
r = tracker.predict(im)
total += time.perf_counter() - start
print(1. / (total / 100.))
sys.exit(0)
target_ip = args.ip
target_port = args.port
if args.faces >= 40:
print("Transmission of tracking data over network is not supported with 40 or more faces.")
def process_video(video_path):
output_logfile = None
if args.log_output != "":
output_logfile = open(args.log_output, "w")
sys.stdout = OutputLog(output_logfile, sys.stdout)
sys.stderr = OutputLog(output_logfile, sys.stderr)
if os.name == 'nt':
import dshowcapture
if args.blackmagic == 1:
dshowcapture.set_bm_enabled(True)
if not args.blackmagic_options is None:
dshowcapture.set_options(args.blackmagic_options)
if not args.priority is None:
import psutil
classes = [psutil.IDLE_PRIORITY_CLASS, psutil.BELOW_NORMAL_PRIORITY_CLASS, psutil.NORMAL_PRIORITY_CLASS, psutil.ABOVE_NORMAL_PRIORITY_CLASS, psutil.HIGH_PRIORITY_CLASS, psutil.REALTIME_PRIORITY_CLASS]
p = psutil.Process(os.getpid())
p.nice(classes[args.priority])
if os.name == 'nt' and (args.list_cameras > 0 or not args.list_dcaps is None):
cap = dshowcapture.DShowCapture()
info = cap.get_info()
unit = 10000000.;
if not args.list_dcaps is None:
formats = {0: "Any", 1: "Unknown", 100: "ARGB", 101: "XRGB", 200: "I420", 201: "NV12", 202: "YV12", 203: "Y800", 300: "YVYU", 301: "YUY2", 302: "UYVY", 303: "HDYC (Unsupported)", 400: "MJPEG", 401: "H264" }
for cam in info:
if args.list_dcaps == -1:
type = ""
if cam['type'] == "Blackmagic":
type = "Blackmagic: "
print(f"{cam['index']}: {type}{cam['name']}")
if args.list_dcaps != -1 and args.list_dcaps != cam['index']:
continue
for caps in cam['caps']:
format = caps['format']
if caps['format'] in formats:
format = formats[caps['format']]
if caps['minCX'] == caps['maxCX'] and caps['minCY'] == caps['maxCY']:
print(f" {caps['id']}: Resolution: {caps['minCX']}x{caps['minCY']} FPS: {unit/caps['maxInterval']:.3f}-{unit/caps['minInterval']:.3f} Format: {format}")
else:
print(f" {caps['id']}: Resolution: {caps['minCX']}x{caps['minCY']}-{caps['maxCX']}x{caps['maxCY']} FPS: {unit/caps['maxInterval']:.3f}-{unit/caps['minInterval']:.3f} Format: {format}")
else:
if args.list_cameras == 1:
print("Available cameras:")
for cam in info:
type = ""
if cam['type'] == "Blackmagic":
type = "Blackmagic: "
if args.list_cameras == 1:
print(f"{cam['index']}: {type}{cam['name']}")
else:
print(f"{type}{cam['name']}")
cap.destroy_capture()
sys.exit(0)
import numpy as np
import time
import cv2
import socket
import struct
import json
from input_reader import InputReader, VideoReader, DShowCaptureReader, try_int
from tracker import Tracker, get_model_base_path
from tqdm import tqdm
if args.benchmark > 0:
model_base_path = get_model_base_path(args.model_dir)
im = cv2.imread(os.path.join(model_base_path, "benchmark.bin"), cv2.IMREAD_COLOR)
results = []
for model_type in [3, 2, 1, 0, -1, -2, -3]:
tracker = Tracker(224, 224, threshold=0.1, max_threads=args.max_threads, max_faces=1, discard_after=0, scan_every=0, silent=True, model_type=model_type, model_dir=args.model_dir, no_gaze=(model_type == -1), detection_threshold=0.1, use_retinaface=0, max_feature_updates=900, static_model=True if args.no_3d_adapt == 1 else False)
tracker.detected = 1
tracker.faces = [(0, 0, 224, 224)]
total = 0.0
for i in range(100):
start = time.perf_counter()
r = tracker.predict(im)
# r = tracker.predict(im)
total += time.perf_counter() - start
print(1. / (total / 100.))
sys.exit(0)
target_ip = args.ip
target_port = args.port
if args.faces >= 40:
print("Transmission of tracking data over network is not supported with 40 or more faces.")
fps = 24
dcap = None
use_dshowcapture_flag = False
if os.name == 'nt':
fps = args.fps
dcap = args.dcap
use_dshowcapture_flag = True if args.use_dshowcapture == 1 else False
input_reader = InputReader(video_path, args.raw_rgb, args.width, args.height, fps, use_dshowcapture=use_dshowcapture_flag, dcap=dcap)
if args.dcap == -1 and type(input_reader) == DShowCaptureReader:
fps = min(fps, input_reader.device.get_fps())
else:
input_reader = InputReader(video_path, args.raw_rgb, args.width, args.height, fps, use_dshowcapture=use_dshowcapture_flag)
# if type(input_reader.reader) == VideoReader:
# fps = 0.0
log = None
out = None
first = True
height = 0
width = 0
tracker = None
sock = None
total_tracking_time = 0.0
tracking_time = 0.0
tracking_frames = 0
framecount = 0
eye_blink_frames = 0
eye_blink_lst = []
eye_blink_temp = []
COUNTER = 0
TOTAL = 0
current_frame = 1
blink_start = 0
blink_end = 0
closeness = 0
output_closeness = []
output_blinks = []
blink_info = (0,0)
processed_frame = []
frame_info_list = []
lStart = 42
lEnd = 48
rStart = 36
rEnd = 42
ear_th = 0.18
consec_th = 3
up_to = None
array_blink_threshold = list()
ear_list = list()
col=['F1',"F2","F3","F4","F5",'F6',"F7", "F8", "F9", "F10", "F11", "F12", "F13"]
features = ["eye_l", "eye_r", "eyebrow_steepness_l", "eyebrow_updown_l", "eyebrow_quirk_l", "eyebrow_steepness_r", "eyebrow_updown_r", "eyebrow_quirk_r", "mouth_corner_updown_l", "mouth_corner_inout_l", "mouth_corner_updown_r", "mouth_corner_inout_r", "mouth_open", "mouth_wide"]
if args.log_data != "":
log = open(args.log_data, "w")
log.write("Frame,Time,Width,Height,FPS,Face,FaceID,RightOpen,LeftOpen,AverageConfidence,Success3D,PnPError,RotationQuat.X,RotationQuat.Y,RotationQuat.Z,RotationQuat.W,Euler.X,Euler.Y,Euler.Z,RVec.X,RVec.Y,RVec.Z,TVec.X,TVec.Y,TVec.Z")
for i in range(66):
log.write(f",Landmark[{i}].X,Landmark[{i}].Y,Landmark[{i}].Confidence")
for i in range(66):
log.write(f",Point3D[{i}].X,Point3D[{i}].Y,Point3D[{i}].Z")
for feature in features:
log.write(f",{feature}")
log.write("\r\n")
log.flush()
is_camera = video_path == str(try_int(video_path))
try:
attempt = 0
frame_time = time.perf_counter()
target_duration = 0
if fps > 0:
target_duration = 1. / float(fps)
repeat = args.repeat_video != 0 and type(input_reader.reader) == VideoReader
need_reinit = 0
failures = 0
source_name = input_reader.name
blink_count = 0
blink_count_origin = 0
while repeat or input_reader.is_open():
if not input_reader.is_open() or need_reinit == 1:
input_reader = InputReader(video_path, args.raw_rgb, args.width, args.height, fps, use_dshowcapture=use_dshowcapture_flag, dcap=dcap)
if input_reader.name != source_name:
print(f"Failed to reinitialize camera and got {input_reader.name} instead of {source_name}.")
sys.exit(1)
need_reinit = 2
time.sleep(0.02)
continue
if not input_reader.is_ready():
time.sleep(0.02)
continue
ret, frame = input_reader.read()
fps = input_reader.get_fps()
frame_count = int(input_reader.get_frame())
duration = frame_count/fps
video_info_dict = {
'fps': fps,
'frame_count': frame_count,
'duration(s)': duration
}
# frame = cv2.flip(frame,1)
#2 -50 - 0.5 -20,-50
# frame = cv2.convertScaleAbs(frame, -1, 0.5, -20)
if not ret:
if repeat:
if need_reinit == 0:
need_reinit = 1
continue
elif is_camera:
attempt += 1
if attempt > 30:
break
else:
time.sleep(0.02)
if attempt == 3:
need_reinit = 1
continue
else:
break;
attempt = 0
need_reinit = 0
# frame_count += 1
now = time.time()
if first:
first = False
height, width, channels = frame.shape
sock = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
tracker = Tracker(width, height, threshold=args.threshold, max_threads=args.max_threads, max_faces=args.faces, discard_after=args.discard_after, scan_every=args.scan_every, silent=False if args.silent == 0 else True, model_type=args.model, model_dir=args.model_dir, no_gaze=False if args.gaze_tracking != 0 and args.model != -1 else True, detection_threshold=args.detection_threshold, use_retinaface=args.scan_retinaface, max_feature_updates=args.max_feature_updates, static_model=True if args.no_3d_adapt == 1 else False, try_hard=args.try_hard == 1)
if not args.video_out is None:
out = cv2.VideoWriter(args.video_out, cv2.VideoWriter_fourcc('F','F','V','1'), args.video_fps, (width * args.video_scale, height * args.video_scale))
try:
inference_start = time.perf_counter()
faces = tracker.predict(frame)
# faces = tracker.predict(frame)
if len(faces) > 0:
inference_time = (time.perf_counter() - inference_start)
total_tracking_time += inference_time
tracking_time += inference_time / len(faces)
tracking_frames += 1
# else:
# ear_list.append(np.nan)
# array_blink_threshold.append(np.nan)
packet = bytearray()
detected = False
r_eye_roi_resize = []
l_eye_roi_resize = []
head_list = []
for face_num, f in enumerate(faces):
f = copy.copy(f)
f.id += args.face_id_offset
if f.eye_blink is None:
f.eye_blink = [1, 1]
right_state = "O" if f.eye_blink[0] > 0.30 else "-"
left_state = "O" if f.eye_blink[1] > 0.30 else "-"
if f.eye_blink[0] < 0.7 or f.eye_blink[1] < 0.7:
eye_blink_frames += 1
if args.silent == 0:
print(f"Confidence[{f.id}]: {f.conf:.4f} / 3D fitting error: {f.pnp_error:.4f} / Eyes: {left_state}, {right_state}")
detected = True
if not f.success:
pts_3d = np.zeros((70, 3), np.float32)
packet.extend(bytearray(struct.pack("d", now)))
packet.extend(bytearray(struct.pack("i", f.id)))
packet.extend(bytearray(struct.pack("f", width)))
packet.extend(bytearray(struct.pack("f", height)))
packet.extend(bytearray(struct.pack("f", f.eye_blink[0])))
packet.extend(bytearray(struct.pack("f", f.eye_blink[1])))
packet.extend(bytearray(struct.pack("B", 1 if f.success else 0)))
packet.extend(bytearray(struct.pack("f", f.pnp_error)))
packet.extend(bytearray(struct.pack("f", f.quaternion[0])))
packet.extend(bytearray(struct.pack("f", f.quaternion[1])))
packet.extend(bytearray(struct.pack("f", f.quaternion[2])))
packet.extend(bytearray(struct.pack("f", f.quaternion[3])))
packet.extend(bytearray(struct.pack("f", f.euler[0])))
packet.extend(bytearray(struct.pack("f", f.euler[1])))
packet.extend(bytearray(struct.pack("f", f.euler[2])))
packet.extend(bytearray(struct.pack("f", f.translation[0])))
packet.extend(bytearray(struct.pack("f", f.translation[1])))
packet.extend(bytearray(struct.pack("f", f.translation[2])))
r_eye = [[f.lms[36][1], f.lms[36][0]], [f.lms[37][1], f.lms[37][0]], [f.lms[38][1], f.lms[38][0]], [f.lms[39][1], f.lms[39][0]], [f.lms[40][1], f.lms[40][0]], [f.lms[41][1], f.lms[41][0]]]
l_eye = [[f.lms[42][1], f.lms[42][0]], [f.lms[43][1], f.lms[43][0]], [f.lms[44][1], f.lms[44][0]], [f.lms[45][1], f.lms[45][0]], [f.lms[46][1], f.lms[46][0]], [f.lms[47][1], f.lms[47][0]]]
l_eye_ratio = eye_aspect_ratio(l_eye)
r_eye_ratio = eye_aspect_ratio(r_eye)
ear = (l_eye_ratio + r_eye_ratio) / 2.0
ear_model = (f.eye_blink[0] + f.eye_blink[1]) / 2.0
if ear_model < 0.7: #0.21
COUNTER += 1
closeness = 1
output_closeness.append(closeness)
else:
if COUNTER >= consec_th:
TOTAL += 1
blink_start = current_frame - COUNTER
blink_end = current_frame - 1
blink_info = (blink_start, blink_end)
output_blinks.append(blink_info)
COUNTER = 0
closeness = 0
output_closeness.append(closeness)
frame_info = {
'frame_no': current_frame,
'face_detected': 1,
'face_coordinates': 0,
'left_eye_coor': 0,
'right_eye_coor': 0,
'left_ear': l_eye_ratio,
'right_ear': r_eye_ratio,
'avg_ear': ear,
'avg_ear_model': ear_model,
'closeness': closeness,
'blink_no': TOTAL,
'blink_start_frame': blink_start,
'blink_end_frame': blink_end,
'reserved_for_calibration': False
}
frame_info_list.append(frame_info)
processed_frame.append(frame)
current_frame += 1
# frame_info_df = pd.DataFrame(frame_info_list) # debug
framecount += 1
if not log is None:
log.write(f"{framecount},{now},{width},{height},{args.fps},{face_num},{f.id},{f.eye_blink[0]},{f.eye_blink[1]},{f.conf},{f.success},{f.pnp_error},{f.quaternion[0]},{f.quaternion[1]},{f.quaternion[2]},{f.quaternion[3]},{f.euler[0]},{f.euler[1]},{f.euler[2]},{f.rotation[0]},{f.rotation[1]},{f.rotation[2]},{f.translation[0]},{f.translation[1]},{f.translation[2]}")
for (x,y,c) in f.lms:
packet.extend(bytearray(struct.pack("f", c)))
if args.visualize > 1:
frame = cv2.putText(frame, str(f.id), (int(f.bbox[0]), int(f.bbox[1])), cv2.FONT_HERSHEY_SIMPLEX, 0.75, (255,0,0))
frame = cv2.putText(frame, "FPS : %0.1f" % fps, (10, 50), cv2.FONT_HERSHEY_SIMPLEX, 0.55, (0,255,0), 1, cv2.LINE_AA)
if args.visualize > 2:
frame = cv2.putText(frame, f"{f.conf:.4f}", (int(f.bbox[0] + 18), int(f.bbox[1] - 6)), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0,0,255))
# cv2.imwrite('frame.jpg', frame)
for pt_num, (x,y,c) in enumerate(f.lms):
packet.extend(bytearray(struct.pack("f", y)))
packet.extend(bytearray(struct.pack("f", x)))
if not log is None:
log.write(f",{y},{x},{c}")
if pt_num == 66 and (f.eye_blink[0] < 0.30 or c < 0.30):
continue
if pt_num == 67 and (f.eye_blink[1] < 0.30 or c < 0.30):
continue
x = int(x + 0.5)
y = int(y + 0.5)
if args.visualize != 0 or not out is None:
if args.visualize > 3:
frame = cv2.putText(frame, str(pt_num), (int(y), int(x)), cv2.FONT_HERSHEY_SIMPLEX, 0.25, (255,255,0))
color = (0, 255, 0)
if pt_num >= 66:
color = (255, 255, 0)
if not (x < 0 or y < 0 or x >= height or y >= width):
frame[int(x), int(y)] = color
x += 1
if not (x < 0 or y < 0 or x >= height or y >= width):
frame[int(x), int(y)] = color
y += 1
if not (x < 0 or y < 0 or x >= height or y >= width):
frame[int(x), int(y)] = color
x -= 1
if not (x < 0 or y < 0 or x >= height or y >= width):
frame[int(x), int(y)] = color
if args.pnp_points != 0 and (args.visualize != 0 or not out is None) and f.rotation is not None:
if args.pnp_points > 1:
projected = cv2.projectPoints(f.face_3d[0:66], f.rotation, f.translation, tracker.camera, tracker.dist_coeffs)
else:
projected = cv2.projectPoints(f.contour, f.rotation, f.translation, tracker.camera, tracker.dist_coeffs)
for [(x,y)] in projected[0]:
x = int(x + 0.5)
y = int(y + 0.5)
if not (x < 0 or y < 0 or x >= height or y >= width):
frame[int(x), int(y)] = (0, 255, 255)
x += 1
if not (x < 0 or y < 0 or x >= height or y >= width):
frame[int(x), int(y)] = (0, 255, 255)
y += 1
if not (x < 0 or y < 0 or x >= height or y >= width):
frame[int(x), int(y)] = (0, 255, 255)
x -= 1
if not (x < 0 or y < 0 or x >= height or y >= width):
frame[int(x), int(y)] = (0, 255, 255)
for (x,y,z) in f.pts_3d:
packet.extend(bytearray(struct.pack("f", x)))
packet.extend(bytearray(struct.pack("f", -y)))
packet.extend(bytearray(struct.pack("f", -z)))
if not log is None:
log.write(f",{x},{-y},{-z}")
if f.current_features is None:
f.current_features = {}
for feature in features:
if not feature in f.current_features:
f.current_features[feature] = 0
packet.extend(bytearray(struct.pack("f", f.current_features[feature])))
if not log is None:
log.write(f",{f.current_features[feature]}")
if not log is None:
log.write("\r\n")
log.flush()
if detected and len(faces) < 40:
sock.sendto(packet, (target_ip, target_port))
if not out is None:
video_frame = frame
if args.video_scale != 1:
video_frame = cv2.resize(frame, (width * args.video_scale, height * args.video_scale), interpolation=cv2.INTER_NEAREST)
out.write(video_frame)
if args.video_scale != 1:
del video_frame
failures = 0
except Exception as e:
if e.__class__ == KeyboardInterrupt:
if args.silent == 0:
print("Quitting")
break
traceback.print_exc()
failures += 1
if failures > 30:
break
collected = False
del frame
duration = time.perf_counter() - frame_time
while duration < target_duration:
if not collected:
gc.collect()
collected = True
duration = time.perf_counter() - frame_time
sleep_time = target_duration - duration
if sleep_time > 0:
time.sleep(sleep_time)
duration = time.perf_counter() - frame_time
frame_time = time.perf_counter()
except KeyboardInterrupt:
if args.silent == 0:
print("Quitting")
input_reader.close()
if not out is None:
out.release()
cv2.destroyAllWindows()
if args.silent == 0 and tracking_frames > 0:
average_tracking_time = 1000 * tracking_time / tracking_frames
print(f"Average tracking time per detected face: {average_tracking_time:.2f} ms")
print(f"Tracking time: {total_tracking_time:.3f} s\nFrames: {tracking_frames}")
frame_info_df = pd.DataFrame(frame_info_list)
frame_info_df['output_closeness'] = output_closeness
file_name = os.path.basename(video_path)
output_str = 'Processing {} has done.\n\n'.format(file_name)
return frame_info_df, output_closeness, output_blinks, processed_frame, video_info_dict, output_str
