File size: 2,594 Bytes
0037215
e73a5e2
0037215
 
 
 
 
e73a5e2
0037215
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e73a5e2
 
 
 
 
0037215
 
 
 
 
 
e73a5e2
 
 
 
 
0037215
 
 
 
e73a5e2
0037215
 
 
 
e73a5e2
 
 
 
0037215
e73a5e2
9159b0a
e73a5e2
 
 
9159b0a
 
0037215
 
 
 
e73a5e2
9159b0a
0037215
 
 
 
 
 
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
import gradio as gr
from transformers import AutoProcessor, AutoModelForCausalLM, MarianMTModel, MarianTokenizer
from PIL import Image
import torch
from gtts import gTTS
import os

# Funções auxiliares
def prepare_image(image_path):
    image = Image.open(image_path).convert("RGB")
    inputs = processor(images=image, return_tensors="pt").to(device)
    return image, inputs.pixel_values

def generate_caption(pixel_values):
    model.eval()
    with torch.no_grad():
        generated_ids = model.generate(
            pixel_values=pixel_values,
            max_length=50,
            num_beams=4,
            early_stopping=True,
            no_repeat_ngram_size=2
        )
    return processor.batch_decode(generated_ids, skip_special_tokens=True)[0]

def translate_to_portuguese(text):
    inputs = translation_tokenizer(text, return_tensors="pt", truncation=True).to(device)
    translated_ids = translation_model.generate(inputs["input_ids"], max_length=50, num_beams=4, early_stopping=True)
    return translation_tokenizer.batch_decode(translated_ids, skip_special_tokens=True)[0]

def text_to_speech_gtts(text, lang='pt'):
    tts = gTTS(text=text, lang=lang)
    tts.save("output.mp3")
    return "output.mp3"

# Carregar os modelos
processor = AutoProcessor.from_pretrained("microsoft/git-base")
model = AutoModelForCausalLM.from_pretrained("microsoft/git-base")
translation_model_name = 'Helsinki-NLP/opus-mt-tc-big-en-pt'
translation_tokenizer = MarianTokenizer.from_pretrained(translation_model_name)
translation_model = MarianMTModel.from_pretrained(translation_model_name)

# Configurar o dispositivo (GPU ou CPU)
device = "cuda" if torch.cuda.is_available() else "cpu"
model.to(device)
translation_model.to(device)

# Função principal para processar a imagem e gerar a voz
def process_image(image):
    _, pixel_values = prepare_image(image)
    caption_en = generate_caption(pixel_values)
    caption_pt = translate_to_portuguese(caption_en)
    audio_file = text_to_speech_gtts(caption_pt)
    return caption_pt, audio_file

# Caminhos para as imagens de exemplo (supondo que estejam no mesmo diretório que o script)
example_image_paths = [
    "example1.png",
    "example2.png",
    "example3.png"
]

# Interface Gradio
iface = gr.Interface(
    fn=process_image,
    inputs=gr.Image(type="filepath"),
    outputs=[gr.Textbox(), gr.Audio(type="filepath")],
    examples=example_image_paths,
    title="Image to Voice",
    description="Gera uma descrição em português e a converte em voz a partir de uma imagem."
)

if __name__ == "__main__":
    iface.launch()