-
Notifications
You must be signed in to change notification settings - Fork 12
/
gen_resource_pkg.py
299 lines (212 loc) · 10.6 KB
/
gen_resource_pkg.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
import os
import cv2
import numpy as np
from tqdm import tqdm
import torch
import pickle
import sys
from config import log, params
import subprocess
import config as cfg
from config import log, params
from facelib.utils.face_restoration_helper import FaceRestoreHelper
from basicsr.utils import img2tensor, tensor2img
from torchvision.transforms.functional import normalize
from basicsr.utils.registry import ARCH_REGISTRY
'''
生成资源包
生成之后直接在params['app']['video_based_path']目录下，每个虚拟人ID一个子目录，子目录里放该人的所有资源包文件
'''
# 直接跑codeformer的保存pkl模块
'''
获取视频帧序列
'''
def get_video_frame_list(face_video_path):
video_stream = cv2.VideoCapture(face_video_path)
fps = int(video_stream.get(cv2.CAP_PROP_FPS))
log.logger.info('Reading video frames...')
full_frames = []
while True:
still_reading, frame = video_stream.read()
if not still_reading:
video_stream.release()
break
full_frames.append(frame)
return full_frames, fps
# 跑 codeformer 的 pkl 转 成 wav2lip认识的pkl
def face_detect_by_codeformer(images, codeformer_det_faces_pkl_path ,video_name, fps, save_pkl_path):
log.logger.info('把codeformer格式的pkl转为wav2lip能接受的')
'''
输入 images 把codeformer输出的pkl转为wav2lip可接受的pkl，使用localpose时可直接读此pkl
'''
with open(codeformer_det_faces_pkl_path, 'rb') as f:
face_data = pickle.load(f)
results = []
pady1, pady2, padx1, padx2 = params['app']['wav2lip']['pads']
for image, bbox in zip(images, face_data):
# import pdb;pdb.set_trace()
y1 = round( max(0, bbox[0][1] - pady1) )
y2 = round( min(image.shape[0], bbox[0][3] + pady2) )
x1 = round( max(0, bbox[0][0] - padx1) )
x2 = round( min(image.shape[1], bbox[0][2] + padx2) )
results.append([x1, y1, x2, y2])
boxes = np.array(results)
# import pdb;pdb.set_trace()
if not params['app']['wav2lip']['nosmooth']: boxes = get_smoothened_boxes(boxes, T=5)
results = [[image[y1: y2, x1:x2], (y1, y2, x1, x2)] for image, (x1, y1, x2, y2) in zip(images, boxes)]
# print("datagen:",results) ## 把这个results写死
# import pdb;pdb.set_trace()
# print("将转成wav2lip格式的list保存到文件")
wav2lip_det_faces_pkl_path = os.path.dirname(codeformer_det_faces_pkl_path)
# print("wav2lip_det_faces_pkl_path:", wav2lip_det_faces_pkl_path)
# import pdb
# pdb.set_trace()
with open(os.path.join(save_pkl_path, "%s.pkl" % video_name), "wb") as f5:
pickle.dump(results, f5)
# import pdb;pdb.set_trace()
return None
'''
获取平滑之后的boxes
'''
def get_smoothened_boxes(boxes, T):
for i in range(len(boxes)):
if i + T > len(boxes):
window = boxes[len(boxes) - T:]
else:
window = boxes[i : i + T]
boxes[i] = np.mean(window, axis=0)
return boxes
def gen_resource_pkg(video_source_path, output_path):
video_name, file_ext = os.path.splitext(os.path.basename(video_source_path))
save_pkl_path = os.path.join(output_path, str(video_name))
os.makedirs(save_pkl_path, exist_ok=True)
imglist, fps = get_video_frame_list(video_source_path)
if fps == 0:
log.logger.error(video_source_path,"导出的 1080 1920 视频的 有问题！！！")
sys.exit(0)
net = ARCH_REGISTRY.get('CodeFormer')(dim_embd=512, codebook_size=1024, n_head=8, n_layers=9,
connect_list=['32', '64', '128', '256']).to(cfg.device)
checkpoint = torch.load("./checkpoint/codeformer.pth")['params_ema']
net.load_state_dict(checkpoint)
net.eval()
face_helper = FaceRestoreHelper(
params['app']['inpainting']['upscale'],
face_size=512,
crop_ratio=(1, 1),
det_model="retinaface_resnet50",
save_ext='png',
use_parse=True,
device=cfg.device) # 人脸关键点检测用的是 retinaface_resnet50
total_all_landmarks_5 = []
total_det_faces = []
total_affine_matrices = []
total_inverse_affine_matrices = []
output_imglist = []
for frame_idx in tqdm(range(len(imglist)), desc='正在逐帧处理视频，获取资源包'):
# print("frame_idx:", frame_idx)
bg_upsampler = None
face_upsampler = None
img = imglist[frame_idx]
face_helper.clean_all()
face_helper.read_image(img)
# import pdb;pdb.set_trace()
num_det_faces = face_helper.get_face_landmarks_5(only_center_face=params['app']['inpainting']['only_center_face'], resize=640, eye_dist_threshold=5)
# align and warp each face
# import pdb;pdb.set_trace()
face_helper.align_warp_face() # 对齐
# face restoration for each cropped face
# print("face_helper.cropped_faces:", face_helper.cropped_faces)
for idx, cropped_face in enumerate(face_helper.cropped_faces):
# prepare data
cropped_face_t = img2tensor(cropped_face / 255., bgr2rgb=True, float32=True)
normalize(cropped_face_t, (0.5, 0.5, 0.5), (0.5, 0.5, 0.5), inplace=True)
cropped_face_t = cropped_face_t.unsqueeze(0).to(cfg.device)
# print("cropped_face_t.shape",cropped_face_t.shape)
try:
with torch.no_grad():
output = net(cropped_face_t, w=params['app']['inpainting']['fidelity_weight'], adain=True)[0] # [1, 3, 512, 512]
output_copy = output.clone()
restored_face = tensor2img(output_copy, rgb2bgr=True, min_max=(-1, 1)) # restored_face [512, 512, 3]
normalize(output, (0.5, 0.5, 0.5), (0.5, 0.5, 0.5), inplace=True)
with torch.no_grad():
out = face_helper.face_parse(output)[0]
out = out.argmax(dim=1).squeeze().cpu().numpy()
# print("out.shape",out.shape)
del output
torch.cuda.empty_cache()
except Exception as e:
log.logger.error(f'\tFailed inference for CodeFormer: %s' % (traceback.format_exc()))
restored_face = tensor2img(cropped_face_t, rgb2bgr=True, min_max=(-1, 1))
restored_face = restored_face.astype('uint8')
face_helper.add_restored_face(restored_face, cropped_face)
# paste_back
if params['app']['inpainting']['has_aligned'] is False:
# upsample the background
if bg_upsampler is not None:
# Now only support RealESRGAN for upsampling background
bg_img = bg_upsampler.enhance(img, outscale=params['app']['inpainting']['upscale'])[0]
else:
bg_img = None
face_helper.get_inverse_affine(None)
# paste each restored face to the input image
# import pdb;pdb.set_trace()
if params['app']['inpainting']['face_upsample'] and face_upsampler is not None:
restored_img = face_helper.paste_faces_to_input_image(face_parse_out=out, upsample_img=bg_img, draw_box=params['app']['inpainting']['draw_box'],
face_upsampler=face_upsampler)
else:
restored_img = face_helper.paste_faces_to_input_image(face_parse_out=out, upsample_img=bg_img, draw_box=params['app']['inpainting']['draw_box'])
output_imglist.append(restored_img)
total_all_landmarks_5.append(face_helper.all_landmarks_5)
total_det_faces.append(face_helper.det_faces)
total_affine_matrices.append(face_helper.affine_matrices)
total_inverse_affine_matrices.append(face_helper.inverse_affine_matrices)
with open(os.path.join(save_pkl_path, "all_landmarks_5.pkl"), 'wb') as f1:
pickle.dump(total_all_landmarks_5, f1)
with open(os.path.join(save_pkl_path, "det_faces.pkl"), 'wb') as f2:
pickle.dump(total_det_faces, f2)
with open(os.path.join(save_pkl_path, "affine_matrices.pkl"), 'wb') as f3:
pickle.dump(total_affine_matrices, f3)
with open(os.path.join(save_pkl_path, "inverse_affine_matrices.pkl"), 'wb') as f4:
pickle.dump(total_inverse_affine_matrices, f4)
log.logger.info("所有资源包 pickle dump 完成")
def change_w_h_25fps(origin_video_path, resize_video_path, fps=25, width=1080, height=1920):
directory = origin_video_path
output_path = resize_video_path
for filename in tqdm(os.listdir(directory), desc='正在处理ffmpeg resize图片'):
if filename.endswith(".mp4"):
input_path_tmp = os.path.join(directory, filename)
output_path_tmp = os.path.join(output_path, filename)
subprocess.call(f"ffmpeg -i '{input_path_tmp}' -s '{width}'x'{height}' -r '{fps}' '{output_path_tmp}' -loglevel quiet",
shell=True)
else:
log.logger.info("文件夹中有非mp4文件: %s，已跳过" % filename)
if __name__ == "__main__":
origin_video_path = "./testdata/video_based/"
use_resize = False # 是否调整分辨率和帧数
#############
## 进行调整分辨率和帧数的操作
##############
# 1-1、调整分辨率及FPS
if (use_resize==True):
resize_video_path = "./testdata/resize/"
os.makedirs(resize_video_path, exist_ok=True)
log.logger.info("开始调整分辨率和帧数")
change_w_h_25fps(origin_video_path, resize_video_path, fps=25)
log.logger.info("分辨率和帧数调整完成")
file_names = os.listdir(resize_video_path)
output_path = resize_video_path
ready_video_path = resize_video_path
else:
# 1-2、不调整分辨率及FPS
log.logger.info("不调整分辨率")
file_names = os.listdir(origin_video_path)
output_path = origin_video_path
ready_video_path = origin_video_path
#############
## 2、进行保存资源包的操作
##############
# print("file_names",file_names)
for file_name in tqdm(file_names, desc='正在处理'+ ready_video_path):
if file_name.endswith('.mp4'):
gen_resource_pkg(os.path.join(ready_video_path, file_name), output_path)
print("资源包pkl保存完成！！")