def __init__(self, prefix, epoch, ctx_id=0):
# print('loading',prefix, epoch)
# ctx = mx.gpu(ctx_id)
# sym, arg_params, aux_params = mx.model.load_checkpoint(prefix, epoch)
# all_layers = sym.get_internals()
# sym = all_layers['fc1_output']
image_size = (112,112)
self.image_size = image_size
# model = mx.mod.Module(symbol=sym, context=ctx, label_names = None)
# model.bind(for_training=False, data_shapes=[('data', (2, 3, image_size[0], image_size[1]))])
# model.set_params(arg_params, aux_params)
# self.model = model
src = np.array([
[30.2946, 51.6963],
[65.5318, 51.5014],
[48.0252, 71.7366],
[33.5493, 92.3655],
[62.7299, 92.2041] ], dtype=np.float32 )
src[:,0] += 8.0
self.src = src
self.face_recognizer = face_common.FaceRecognizer(
True,
"/mnt/hdd/PycharmProjects/insightface/detection/scrfd/scrfd_34g_n1/scrfd_34g_shape320x320.onnx",
320, 0.01, 0.4,
True,
"models/backbone.onnx",
0
)
num_face_landmark)
return nme_loss
data_path = "/mnt/hdd/IJB"
target = 'IJBC'
img_path = os.path.join(data_path, './%s/loose_crop' % target)
img_list_path = os.path.join(
data_path, './%s/meta/%s_name_5pts_score.txt' % (target, target.lower()))
img_list = open(img_list_path)
files = img_list.readlines()
print('files:', len(files))
face_recognizer = face_common.FaceRecognizer(
True, "model/retinaface_resnet50_480x480.onnx", 480, 0.01, 0.4, False, "",
0)
result = 0
for img_index, each_line in enumerate(files):
if img_index % 500 == 0:
print('processing', img_index)
name_lmk_score = each_line.strip().split(' ')
img_name = os.path.join(img_path, name_lmk_score[0])
img = cv2.imread(img_name)
target_lmk = np.array([float(x) for x in name_lmk_score[1:-1]],
dtype=np.float32)
target_lmk = target_lmk.reshape((5, 2))
detection = face_recognizer.Detect(img, False, False)
landmarks = detection[0].landmarks
predict_lanmark = []
from __future__ import print_function
import os
import cv2
import torch
import face_common
import numpy as np
save_folder = "./prediction/"
dataset_folder = "../widerface_val/images/"
if __name__ == '__main__':
face_recognizer = face_common.FaceRecognizer(
True, "model/fd_mobilenet_origin.onnx", 480, 0.02, False, "")
torch.set_grad_enabled(False)
testset_folder = dataset_folder
testset_list = dataset_folder[:-7] + "wider_val.txt"
with open(testset_list, 'r') as fr:
test_dataset = fr.read().split()
num_images = len(test_dataset)
for i, img_name in enumerate(test_dataset):
############################# Add face detection here#######################################
image_path = testset_folder + img_name
img = cv2.imread(image_path, cv2.IMREAD_COLOR)
h, w = img.shape[:2]
results = face_recognizer.Detect(img, False, False)
dets = []
for result in results:
dets.append(
[result.x1, result.y1, result.x2, result.y2, result.confident])
dets = np.array(dets)
############################################################################################
from __future__ import print_function
import os
import cv2
import face_common
import numpy as np
save_folder = "./prediction/"
dataset_folder = "../widerface_val/images/"
if __name__ == '__main__':
face_recognizer = face_common.FaceRecognizer(
True, "model/yolov5s-face640x640.onnx", 640, 0.02, 0.5, False, "")
testset_folder = dataset_folder
testset_list = dataset_folder[:-7] + "wider_val.txt"
with open(testset_list, 'r') as fr:
test_dataset = fr.read().split()
num_images = len(test_dataset)
for i, img_name in enumerate(test_dataset):
############################# Add face detection here#######################################
image_path = testset_folder + img_name
img = cv2.imread(image_path, cv2.IMREAD_COLOR)
h, w = img.shape[:2]
results = face_recognizer.Detect(img, False, False)
dets = []
for result in results:
dets.append(
[result.x1, result.y1, result.x2, result.y2, result.confident])
dets = np.array(dets)
############################################################################################
save_name = save_folder + img_name[:-4] + ".txt"
dirname = os.path.dirname(save_name)
if not os.path.isdir(dirname):
from __future__ import print_function
import os
import cv2
import face_common
import numpy as np
save_folder = "./prediction/"
dataset_folder = "../widerface_val/images/"
if __name__ == '__main__':
face_recognizer = face_common.FaceRecognizer(True,
"model/scrfd_10g_bnkps.onnx",
640, 0.02, False, "")
testset_folder = dataset_folder
testset_list = dataset_folder[:-7] + "wider_val.txt"
with open(testset_list, 'r') as fr:
test_dataset = fr.read().split()
num_images = len(test_dataset)
for i, img_name in enumerate(test_dataset):
############################# Add face detection here#######################################
image_path = testset_folder + img_name
img = cv2.imread(image_path, cv2.IMREAD_COLOR)
h, w = img.shape[:2]
results = face_recognizer.Detect(img, False, False)
dets = []
for result in results:
dets.append(
[result.x1, result.y1, result.x2, result.y2, result.confident])
dets = np.array(dets)
############################################################################################
save_name = save_folder + img_name[:-4] + ".txt"
dirname = os.path.dirname(save_name)
def NMELoss(predicted_landmark, target_landmark):
# landmark is a numpy array which has shape [5, 2]
num_face_landmark = 5
leye_nouse_vec = torch.from_numpy(target_landmark[0] - target_landmark[2])
reye_nouse_vec = torch.from_numpy(target_landmark[1] - target_landmark[2])
inter_occular_distance = LA.norm(leye_nouse_vec) + LA.norm(reye_nouse_vec)
loss = nn.MSELoss(reduction="sum")
preloss = loss(torch.from_numpy(predicted_landmark), torch.from_numpy(target_landmark))
nme_loss = torch.sqrt(preloss) / (inter_occular_distance * num_face_landmark)
return nme_loss
face_recognizer = face_common.FaceRecognizer(
True,
"/mnt/hdd/PycharmProjects/Pytorch_Retinaface/weights/model_origin.onnx",
320, 0.01, 0.4,
False,
"",
0
)
result = 0
number_faces = 0
for i in range(len(dataset)):
img, target = dataset[i]
for face in target:
face = face.astype(np.int32)
area = np.abs(face[2] - face[0]) * np.abs(face[3] - face[1])
if face[-1] != -1 and area > 12544:
number_faces = number_faces + 1
target_lmk = face[4:14].astype(np.float32)