File size: 1,557 Bytes
e665f54 b01b23a e665f54 b01b23a e665f54 b01b23a e665f54 b01b23a e665f54 b01b23a e665f54 b01b23a e665f54 b01b23a |
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 |
import os
import gradio as gr
from transformers import TrOCRProcessor, TrOCRForConditionalGeneration
from PIL import Image
import torch
# ?? Chargement du modele et du processor
model_name = "microsoft/trocr-base-handwritten"
model = TrOCRForConditionalGeneration.from_pretrained(model_name)
processor = TrOCRProcessor.from_pretrained(model_name)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)
model.eval()
# ?? Fonction OCR
def ocr_from_image(image_file, ocr_type):
if image_file is None:
return "Veuillez importer une image."
# Pretraitement de l'image
image = Image.open(image_file.name).convert("RGB")
pixel_values = processor(images=image, return_tensors="pt").pixel_values.to(device)
# Generation de texte
with torch.no_grad():
generated_ids = model.generate(pixel_values)
# Decodage du texte genere
generated_text = processor.tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
return generated_text
# ?? Types d'OCR (juste pour l'interface ici)
ocr_types = ["ocr", "format"]
# ?? Interface Gradio
iface = gr.Interface(
fn=ocr_from_image,
inputs=[
gr.File(label="Importer une image", file_types=[".jpg", ".jpeg", ".png"]),
gr.Radio(ocr_types, label="Type d'OCR", value="ocr")
],
outputs="text",
title="?? OCR manuscrit avec TrOCR",
description="Importez une image manuscrite pour extraire le texte avec le modele Microsoft TrOCR."
)
# ?? Lancement
if __name__ == "__main__":
iface.launch()
|