Spaces:
Sleeping
Sleeping
File size: 2,690 Bytes
1a7973f a435eea cce1817 7d7172f 30bd11b a435eea 7d7172f 30bd11b a435eea 30bd11b 7d7172f 30bd11b a435eea 30bd11b a435eea 30bd11b a435eea 30bd11b a435eea 30bd11b 7d7172f 30bd11b |
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 |
import gradio as gr
from transformers import AutoProcessor, BlipForConditionalGeneration, AutoTokenizer,SpeechT5Processor, SpeechT5ForTextToSpeech, SpeechT5HifiGan
import librosa
import numpy as np
import torch
import image_text_model
import audio_model
import open_clip
#CONSTANTS
# Carga el modelo de clasificaci贸n de imagen a texto
blip_processor_large = AutoProcessor.from_pretrained("Salesforce/blip-image-captioning-large")
blip_model_large = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-large")
device = "cuda" if torch.cuda.is_available() else "cpu"
blip_model_large.to(device)
##### IMAGE MODEL TO TEXT, MODEL 1
def generate_caption(processor, model, image, tokenizer=None, use_float_16=False):
inputs = processor(images=image, return_tensors="pt").to(device)
if use_float_16:
inputs = inputs.to(torch.float16)
generated_ids = model.generate(pixel_values=inputs.pixel_values, max_length=50)
if tokenizer is not None:
generated_caption = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
else:
generated_caption = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
return generated_caption
def generate_caption_coca(model, transform, image):
im = transform(image).unsqueeze(0).to(device)
with torch.no_grad(), torch.cuda.amp.autocast():
generated = model.generate(im, seq_len=20)
return open_clip.decode(generated[0].detach()).split("<end_of_text>")[0].replace("<start_of_text>", "")
def generate_captions_speech(image):
caption_blip_large = generate_caption(blip_processor_large, blip_model_large, image)
print('generate_captions>>>'+caption_blip_large)
return caption_blip_large,text_to_speech(caption_blip_large,"Surprise Me!")
#####END IMAGE MODEL TO TEXT
# Define la interfaz de usuario utilizando Gradio entradas y salidas
inputsImg = [
gr.Image(type="pil", label="Imagen"),
]
#Salidas es lo que genera de tetxo y el audio
outputs = [ gr.Textbox(label="Caption generated by BLIP-large"),gr.Audio(type="numpy",label='Transcripcion')]
title = "Clasificaci贸n de imagen a texto y conversi贸n de texto a voz"
description = "Carga una imagen y obt茅n una descripci贸n de texto de lo que contiene la imagen, as铆 como un archivo de audio de la trasncripcion de la imagen en audio descrito."
examples = []
interface = gr.Interface(fn=generate_captions_speech,
inputs=inputsImg,
outputs=outputs,
examples=examples,
title=title,
description=description)
interface.launch() |