def main(args, color=(224, 255, 255)):
fd = service.UltraLightFaceDetecion("weights/RFB-320.tflite",
conf_threshold=0.95)
if args.mode in ["sparse", "pose"]:
fa = service.DepthFacialLandmarks("weights/sparse_face.tflite")
else:
fa = service.DenseFaceReconstruction("weights/dense_face.tflite")
if args.mode == "mesh":
color = service.TrianglesMeshRender("asset/render.so",
"asset/triangles.npy")
handler = getattr(service, args.mode)
cap = cv2.VideoCapture(args.filepath)
while True:
ret, frame = cap.read()
if not ret:
break
# face detection
boxes, scores = fd.inference(frame)
# raw copy for reconstruction
feed = frame.copy()
for results in fa.get_landmarks(feed, boxes):
handler(frame, results, color)
# cv2.imwrite(f'draft/gif/trans/img{counter:0>4}.jpg', frame)
cv2.imshow("demo", frame)
if cv2.waitKey(1) == ord("q"):
break
def main(args, color=(224, 255, 255)):
fd = service.UltraLightFaceDetecion("weights/RFB-320.tflite",
conf_threshold=0.92)
if args.mode in ["sparse", "pose"]:
fa = service.DepthFacialLandmarks("weights/sparse_face.tflite")
else:
fa = service.DenseFaceReconstruction("weights/dense_face.tflite")
if args.mode == "mesh":
mr = service.TrianglesMeshRender("asset/render.so",
"asset/triangles.npy")
cap = cv2.VideoCapture(args.filepath)
while True:
ret, frame = cap.read()
if not ret:
break
# face detection
boxes, scores = fd.inference(frame)
# raw copy for reconstruction
feed = frame.copy()
for landmarks, pose in fa.get_landmarks(feed, boxes):
if args.mode == "sparse":
landmarks = np.round(landmarks).astype(np.int)
for p in landmarks:
cv2.circle(frame, tuple(p), 2, color, 0, cv2.LINE_AA)
draw_poly(frame, landmarks, color=color)
elif args.mode == "dense":
landmarks = np.round(landmarks).astype(np.int)
landmarks = landmarks[:2].T
for p in landmarks[::6]:
cv2.circle(frame, tuple(p), 1, color, 0, cv2.LINE_AA)
elif args.mode == "mesh":
landmarks = landmarks.astype(np.float32)
mr.render(landmarks.T.copy(), frame)
cv2.imshow("result", frame)
if cv2.waitKey(1) == ord('q'):
break
#!/usr/bin/python3
# -*- coding:utf-8 -*-
import numpy as np
import sys
import cv2
import service
import time
fd = service.UltraLightFaceDetecion("weights/RFB-320.tflite",
conf_threshold=0.92)
fa = service.DenseFaceReconstruction("weights/dense_face.tflite")
mr = service.TrianglesMeshRender("asset/render.so", "asset/triangles.npy")
cap = cv2.VideoCapture(sys.argv[1])
counter, rate = 0, cap.get(5)
while True:
ret, frame = cap.read()
if not ret:
break
# face detection
start_time = time.perf_counter()
boxes, scores = fd.inference(frame)
detect_cost = time.perf_counter() - start_time
# raw copy for reconstruction
feed = frame.copy()