forked from minguinho26/Prefix_AAC_ICASSP2023
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Inference.py
60 lines (43 loc) · 1.52 KB
/
Inference.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
import torch
import sys
# custom
from util import *
from AAC_Prefix.AAC_Prefix import * # network
from Train import *
TEST_BATCH_SIZE = 5
USE_CUDA = torch.cuda.is_available()
device = torch.device('cuda:0' if USE_CUDA else 'cpu')
argv_num = 1 + 3
if len(sys.argv) != argv_num :
print("you should write 'table_num', 'setting_num' and 'audio file path'!")
exit()
table_num = sys.argv[1]
setting_num = sys.argv[2]
audio_file_path = sys.argv[3]
# table_num = 1 : Evaluation on Clotho
# table_num = 2 : Evaluation on AudioCaps
# setting_num = 1 : train dataset == test dataset
# setting_num = 2 : train dataset != test dataset
# setting_num = 3 : overall datasets(Clotho & AudioCaps) <- need to test by using compressed audio
if setting_num == 3 :
is_settingnum_3 = True
else :
is_settingnum_3 = False
model = get_model_in_table(table_num, setting_num, device)
# prepare audio input=========
SAMPLE_RATE = 16000
set_length = 30
audio_file, _ = torchaudio.load(audio_file_path)
# slicing
if audio_file.shape[0] > (SAMPLE_RATE * set_length) :
audio_file = audio_file[:SAMPLE_RATE * set_length]
# zero padding
if audio_file.shape[0] < (SAMPLE_RATE * set_length) :
pad_len = (SAMPLE_RATE * set_length) - audio_file.shape[0]
pad_val = torch.zeros(pad_len)
audio_file = torch.cat((audio_file, pad_val), dim=0)
# prepare audio input=========
if len(audio_file.size()) == 3 :
audio_file = audio_file.unsqueeze(0)
pred_caption = model(audio_file, None, beam_search = True)[0][0]
print("Caption :", pred_caption)