Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
|
@@ -2,93 +2,57 @@ import gradio as gr
|
|
| 2 |
import easyocr
|
| 3 |
from transformers import pipeline
|
| 4 |
import numpy as np
|
|
|
|
| 5 |
from docx import Document
|
| 6 |
from pptx import Presentation
|
| 7 |
import os
|
| 8 |
-
from PIL import Image
|
| 9 |
|
| 10 |
-
#
|
| 11 |
-
reader = easyocr.Reader([
|
| 12 |
|
| 13 |
-
#
|
| 14 |
summarizer = pipeline("summarization", model="facebook/bart-large-cnn")
|
| 15 |
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
Extraire le texte d'une image avec EasyOCR et le rΓ©sumer avec BART
|
| 19 |
-
|
| 20 |
-
Args:
|
| 21 |
-
image: image PIL ou tableau numpy
|
| 22 |
-
|
| 23 |
-
Returns:
|
| 24 |
-
tuple: (texte_extrait, texte_resume)
|
| 25 |
-
"""
|
| 26 |
try:
|
| 27 |
-
# Convertir l'image PIL en numpy si nΓ©cessaire
|
| 28 |
if isinstance(image, Image.Image):
|
| 29 |
image = np.array(image)
|
| 30 |
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
# Combiner le texte extrait
|
| 35 |
-
texte_extrait = " ".join([result[1] for result in results])
|
| 36 |
-
|
| 37 |
-
if not texte_extrait.strip():
|
| 38 |
-
return "Aucun texte dΓ©tectΓ© dans l'image.", "Aucun texte Γ rΓ©sumer."
|
| 39 |
-
|
| 40 |
-
# VΓ©rifier le nombre de mots
|
| 41 |
-
nombre_mots = len(texte_extrait.split())
|
| 42 |
-
|
| 43 |
-
if nombre_mots < 30:
|
| 44 |
-
return texte_extrait, "Le texte est trop court pour Γͺtre rΓ©sumΓ© (minimum 30 mots)."
|
| 45 |
-
|
| 46 |
-
# Paramètres du résumé
|
| 47 |
-
max_length = min(150, nombre_mots)
|
| 48 |
-
min_length = min(30, nombre_mots // 2)
|
| 49 |
-
|
| 50 |
-
# RΓ©sumΓ©
|
| 51 |
-
resume = summarizer(
|
| 52 |
-
if not text or len(text.split()) < 30:
|
| 53 |
-
return "Texte trop court pour Γͺtre rΓ©sumΓ©."
|
| 54 |
|
| 55 |
-
|
| 56 |
-
|
| 57 |
|
| 58 |
-
|
| 59 |
-
|
| 60 |
-
max_length=max_length,
|
| 61 |
-
min_length=min_length,
|
| 62 |
-
do_sample=False
|
| 63 |
-
)
|
| 64 |
-
|
| 65 |
-
return summary[0]["summary_text"]
|
| 66 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 67 |
)
|
| 68 |
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
return texte_extrait, texte_resume
|
| 72 |
|
| 73 |
except Exception as e:
|
| 74 |
-
return f"Erreur : {
|
| 75 |
-
|
| 76 |
-
|
| 77 |
|
|
|
|
| 78 |
def extract_text_from_file(file_path):
|
| 79 |
ext = os.path.splitext(file_path)[1].lower()
|
| 80 |
|
| 81 |
-
# TXT
|
| 82 |
if ext == ".txt":
|
| 83 |
with open(file_path, "r", encoding="utf-8", errors="ignore") as f:
|
| 84 |
return f.read()
|
| 85 |
|
| 86 |
-
# WORD
|
| 87 |
elif ext == ".docx":
|
| 88 |
doc = Document(file_path)
|
| 89 |
-
return "\n".join(
|
| 90 |
|
| 91 |
-
# POWERPOINT
|
| 92 |
elif ext == ".pptx":
|
| 93 |
prs = Presentation(file_path)
|
| 94 |
text = []
|
|
@@ -101,44 +65,67 @@ def extract_text_from_file(file_path):
|
|
| 101 |
else:
|
| 102 |
return ""
|
| 103 |
|
| 104 |
-
|
| 105 |
-
|
| 106 |
-
|
| 107 |
-
|
| 108 |
-
|
| 109 |
-
|
| 110 |
-
2. Il est rΓ©sumΓ© avec une IA (BART)
|
| 111 |
|
| 112 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 113 |
""")
|
| 114 |
|
| 115 |
-
with gr.
|
| 116 |
-
with gr.Column():
|
| 117 |
-
file_input = gr.File(
|
| 118 |
-
label="TΓ©lΓ©verser un fichier (TXT, DOCX, PPTX)",
|
| 119 |
-
file_types=[".txt", ".docx", ".pptx"]
|
| 120 |
-
)
|
| 121 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 122 |
)
|
| 123 |
-
submit_btn = gr.Button("Extraire et rΓ©sumer", variant="primary")
|
| 124 |
|
| 125 |
-
|
| 126 |
-
|
| 127 |
-
|
| 128 |
-
|
|
|
|
| 129 |
)
|
| 130 |
-
|
| 131 |
-
|
| 132 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 133 |
)
|
| 134 |
|
| 135 |
-
submit_btn.click(
|
| 136 |
-
fn=process_file,
|
| 137 |
-
inputs=file_input,
|
| 138 |
-
outputs=[sortie_texte, sortie_resume]
|
| 139 |
-
)
|
| 140 |
-
|
| 141 |
-
|
| 142 |
-
# Lancer l'application
|
| 143 |
if __name__ == "__main__":
|
| 144 |
demo.launch()
|
|
|
|
| 2 |
import easyocr
|
| 3 |
from transformers import pipeline
|
| 4 |
import numpy as np
|
| 5 |
+
from PIL import Image
|
| 6 |
from docx import Document
|
| 7 |
from pptx import Presentation
|
| 8 |
import os
|
|
|
|
| 9 |
|
| 10 |
+
# OCR (fr + en)
|
| 11 |
+
reader = easyocr.Reader(["fr", "en"], gpu=False)
|
| 12 |
|
| 13 |
+
# RΓ©sumΓ© BART
|
| 14 |
summarizer = pipeline("summarization", model="facebook/bart-large-cnn")
|
| 15 |
|
| 16 |
+
# ---------- IMAGE (OCR) ----------
|
| 17 |
+
def process_image(image):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 18 |
try:
|
|
|
|
| 19 |
if isinstance(image, Image.Image):
|
| 20 |
image = np.array(image)
|
| 21 |
|
| 22 |
+
results = reader.readtext(image, paragraph=True)
|
| 23 |
+
text = " ".join([r[1] for r in results])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 24 |
|
| 25 |
+
if not text.strip():
|
| 26 |
+
return "Aucun texte dΓ©tectΓ©.", "RΓ©sumΓ© impossible."
|
| 27 |
|
| 28 |
+
if len(text.split()) < 30:
|
| 29 |
+
return text, "Texte trop court pour Γͺtre rΓ©sumΓ©."
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 30 |
|
| 31 |
+
wc = len(text.split())
|
| 32 |
+
summary = summarizer(
|
| 33 |
+
text,
|
| 34 |
+
max_length=min(150, wc),
|
| 35 |
+
min_length=min(30, wc // 2),
|
| 36 |
+
do_sample=False
|
| 37 |
)
|
| 38 |
|
| 39 |
+
return text, summary[0]["summary_text"]
|
|
|
|
|
|
|
| 40 |
|
| 41 |
except Exception as e:
|
| 42 |
+
return f"Erreur : {e}", "Erreur."
|
|
|
|
|
|
|
| 43 |
|
| 44 |
+
# ---------- FICHIER ----------
|
| 45 |
def extract_text_from_file(file_path):
|
| 46 |
ext = os.path.splitext(file_path)[1].lower()
|
| 47 |
|
|
|
|
| 48 |
if ext == ".txt":
|
| 49 |
with open(file_path, "r", encoding="utf-8", errors="ignore") as f:
|
| 50 |
return f.read()
|
| 51 |
|
|
|
|
| 52 |
elif ext == ".docx":
|
| 53 |
doc = Document(file_path)
|
| 54 |
+
return "\n".join(p.text for p in doc.paragraphs)
|
| 55 |
|
|
|
|
| 56 |
elif ext == ".pptx":
|
| 57 |
prs = Presentation(file_path)
|
| 58 |
text = []
|
|
|
|
| 65 |
else:
|
| 66 |
return ""
|
| 67 |
|
| 68 |
+
def process_file(file):
|
| 69 |
+
try:
|
| 70 |
+
text = extract_text_from_file(file.name)
|
| 71 |
+
|
| 72 |
+
if not text.strip():
|
| 73 |
+
return "Aucun texte dΓ©tectΓ©.", "RΓ©sumΓ© impossible."
|
|
|
|
| 74 |
|
| 75 |
+
if len(text.split()) < 30:
|
| 76 |
+
return text, "Texte trop court pour Γͺtre rΓ©sumΓ©."
|
| 77 |
+
|
| 78 |
+
wc = len(text.split())
|
| 79 |
+
summary = summarizer(
|
| 80 |
+
text,
|
| 81 |
+
max_length=min(150, wc),
|
| 82 |
+
min_length=min(30, wc // 2),
|
| 83 |
+
do_sample=False
|
| 84 |
+
)
|
| 85 |
+
|
| 86 |
+
return text, summary[0]["summary_text"]
|
| 87 |
+
|
| 88 |
+
except Exception as e:
|
| 89 |
+
return f"Erreur : {e}", "Erreur."
|
| 90 |
+
|
| 91 |
+
# ---------- INTERFACE ----------
|
| 92 |
+
with gr.Blocks(title="OCR & RΓ©sumΓ© (Image + Fichier)") as demo:
|
| 93 |
+
gr.Markdown("""
|
| 94 |
+
# π§ OCR & RΓ©sumΓ© intelligent
|
| 95 |
+
- πΌοΈ Images : OCR + rΓ©sumΓ©
|
| 96 |
+
- π Fichiers : TXT / DOCX / PPTX
|
| 97 |
""")
|
| 98 |
|
| 99 |
+
with gr.Tabs():
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 100 |
|
| 101 |
+
# ----- TAB IMAGE -----
|
| 102 |
+
with gr.Tab("πΌοΈ Image (OCR)"):
|
| 103 |
+
img_input = gr.Image(type="pil", label="TΓ©lΓ©verser une image")
|
| 104 |
+
img_btn = gr.Button("Extraire & RΓ©sumer", variant="primary")
|
| 105 |
+
img_text = gr.Textbox(label="Texte extrait", lines=10)
|
| 106 |
+
img_summary = gr.Textbox(label="RΓ©sumΓ©", lines=5)
|
| 107 |
+
|
| 108 |
+
img_btn.click(
|
| 109 |
+
process_image,
|
| 110 |
+
inputs=img_input,
|
| 111 |
+
outputs=[img_text, img_summary]
|
| 112 |
)
|
|
|
|
| 113 |
|
| 114 |
+
# ----- TAB FILE -----
|
| 115 |
+
with gr.Tab("π Fichier"):
|
| 116 |
+
file_input = gr.File(
|
| 117 |
+
label="TΓ©lΓ©verser un fichier",
|
| 118 |
+
file_types=[".txt", ".docx", ".pptx"]
|
| 119 |
)
|
| 120 |
+
file_btn = gr.Button("Extraire & RΓ©sumer", variant="primary")
|
| 121 |
+
file_text = gr.Textbox(label="Texte extrait", lines=10)
|
| 122 |
+
file_summary = gr.Textbox(label="RΓ©sumΓ©", lines=5)
|
| 123 |
+
|
| 124 |
+
file_btn.click(
|
| 125 |
+
process_file,
|
| 126 |
+
inputs=file_input,
|
| 127 |
+
outputs=[file_text, file_summary]
|
| 128 |
)
|
| 129 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 130 |
if __name__ == "__main__":
|
| 131 |
demo.launch()
|