forked from HengruiLou/CHN-DF
/
eval_main.py
38 lines (33 loc) · 1.82 KB
/
eval_main.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
import argparse
from Unimodal.Eval import *
import os
parser = argparse.ArgumentParser(description='PyTorch CONTINUAL LEARNING')
parser.add_argument('--model', '-m',type=str, default="LIPS", help='model name = [MESO4, MESOINCEPTION4, XCEPTION, EADPOSE]')#TO BE MODIFIED
parser.add_argument('--path_video_model', '-vm',type=str, default="", help='path of the video model')
parser.add_argument('--path_audio_model', '-am',type=str, default="", help='path of the audio model')
parser.add_argument('--path_audio', '-a',type=str, default="/media/data1/mhkim/FakeAVCeleb_PREPROCESSED/SPECTROGRAM/B/TEST", help="path of spectogram (audio)") #TO BE MODIFIED
parser.add_argument('--path_video', '-v',type=str, default="/media/data1/mhkim/FakeAVCeleb_PREPROCESSED/FRAMES_PNG/C/TEST", help='path of path of frame (video)')#TO BE MODIFIED
parser.add_argument('--batch_size', '-nb', type=int, default=200, help='batch size for evaluation')
parser.add_argument('--num_gpu', '-ng', type=str, default='3', help='excuted gpu number')
args = parser.parse_args()
print('GPU num is' , args.num_gpu)
os.environ['CUDA_VISIBLE_DEVICES'] =str(args.num_gpu)
MODEL = args.model
if not args.path_video_model : args.path_video_model = f'./best_{MODEL}_VIDEO.pt'
args.path_audio_model = f'./best_{MODEL}_AUDIO.pt'
if MODEL == 'MESO4':
Eval_MesoNet.Eval(args)
elif MODEL == 'MESOINCEPTION4':
Eval_MesoInceptionNet.Eval(args)
elif MODEL == 'XCEPTION' or MODEL == 'XRAY' or MODEL == 'LIPS':
if MODEL == 'LIPS':
args.path_video = args.path_video.replace('FRAMES_PNG','FRAMES_LIPS_PNG')
if MODEL == 'XRAY':
args.path_video = args.path_video.replace('FRAMES_PNG','FRAMES_XRAY_PNG')
Eval_Xception.Eval(args)
elif MODEL == 'VGG':
Eval_VGG16.Eval(args)
elif MODEL == 'EFFICIENTB0':
Eval_EfficientB0.Eval(args)
elif MODEL == 'F3NET':
Eval_F3Net.Eval(args)