Update app.py

app.py

@@ -1,130 +1,156 @@
 import gradio as gr
 import easyocr
-from transformers import pipeline
 import numpy as np
 from PIL import Image
+from transformers import pipeline
 from docx import Document
 from pptx import Presentation
+from spellchecker import SpellChecker
+from langdetect import detect
+import fitz  # PyMuPDF
 import os
+import tempfile
+
+# -------- OCR --------
+reader = easyocr.Reader(["fr", "en", "ar"], gpu=False)
+
+# -------- Résumés par langue --------
+summarizers = {
+    "fr": pipeline("summarization", model="facebook/bart-large-cnn"),
+    "en": pipeline("summarization", model="facebook/bart-large-cnn"),
+    "ar": pipeline("summarization", model="csebuetnlp/mT5_multilingual_XLSum")
+}
+
+# -------- Correcteur OCR --------
+spell = SpellChecker(language="fr")
+
+def correct_ocr(text):
+    words = text.split()
+    corrected = []
+    for w in words:
+        if w.isalpha():
+            corrected.append(spell.correction(w) or w)
+        else:
+            corrected.append(w)
+    return " ".join(corrected)
+
+# -------- Résumé --------
+def summarize_text(text, lang):
+    wc = len(text.split())
+    if wc < 30:
+        return "Texte trop court pour être résumé."
+
+    model = summarizers.get(lang, summarizers["fr"])
+    result = model(
+        text,
+        max_length=min(150, wc),
+        min_length=min(30, wc // 2),
+        do_sample=False
+    )
+    return result[0]["summary_text"]
+
+# -------- IMAGE --------
+def process_image(image, lang):
+    image = np.array(image)
+    results = reader.readtext(image, paragraph=True)
+    text = " ".join([r[1] for r in results])
+
+    if not text.strip():
+        return "Aucun texte détecté.", "", None
+
+    corrected = correct_ocr(text)
+    summary = summarize_text(corrected, lang)
+
+    file_path = save_summary(summary)
+    return corrected, summary, file_path
+
+# -------- PDF --------
+def extract_text_pdf(path):
+    doc = fitz.open(path)
+    return "\n".join(page.get_text() for page in doc)
+
+# -------- FICHIER --------
+def extract_text(file_path):
+    ext = os.path.splitext(file_path)[1].lower()
 
-# OCR (fr + en)
-reader = easyocr.Reader(["fr", "en"], gpu=False)
-
-# Résumé BART
-summarizer = pipeline("summarization", model="facebook/bart-large-cnn")
-
-# ---------- IMAGE (OCR) ----------
-def process_image(image):
-    try:
-        if isinstance(image, Image.Image):
-            image = np.array(image)
-
-        results = reader.readtext(image, paragraph=True)
-        text = " ".join([r[1] for r in results])
-
-        if not text.strip():
-            return "Aucun texte détecté.", "Résumé impossible."
+    if ext == ".txt":
+        return open(file_path, encoding="utf-8", errors="ignore").read()
 
-        if len(text.split()) < 30:
-            return text, "Texte trop court pour être résumé."
+    if ext == ".docx":
+        return "\n".join(p.text for p in Document(file_path).paragraphs)
 
-        wc = len(text.split())
-        summary = summarizer(
-            text,
-            max_length=min(150, wc),
-            min_length=min(30, wc // 2),
-            do_sample=False
+    if ext == ".pptx":
+        prs = Presentation(file_path)
+        return "\n".join(
+            shape.text for slide in prs.slides for shape in slide.shapes if hasattr(shape, "text")
         )
 
-        return text, summary[0]["summary_text"]
-
-    except Exception as e:
-        return f"Erreur : {e}", "Erreur."
+    if ext == ".pdf":
+        return extract_text_pdf(file_path)
 
-# ---------- FICHIER ----------
-def extract_text_from_file(file_path):
-    ext = os.path.splitext(file_path)[1].lower()
+    return ""
 
-    if ext == ".txt":
-        with open(file_path, "r", encoding="utf-8", errors="ignore") as f:
-            return f.read()
+def process_file(file, lang):
+    text = extract_text(file.name)
 
-    elif ext == ".docx":
-        doc = Document(file_path)
-        return "\n".join(p.text for p in doc.paragraphs)
+    if not text.strip():
+        return "Aucun texte détecté.", "", None
 
-    elif ext == ".pptx":
-        prs = Presentation(file_path)
-        text = []
-        for slide in prs.slides:
-            for shape in slide.shapes:
-                if hasattr(shape, "text"):
-                    text.append(shape.text)
-        return "\n".join(text)
-
-    else:
-        return ""
-
-def process_file(file):
-    try:
-        text = extract_text_from_file(file.name)
-
-        if not text.strip():
-            return "Aucun texte détecté.", "Résumé impossible."
-
-        if len(text.split()) < 30:
-            return text, "Texte trop court pour être résumé."
-
-        wc = len(text.split())
-        summary = summarizer(
-            text,
-            max_length=min(150, wc),
-            min_length=min(30, wc // 2),
-            do_sample=False
-        )
+    corrected = correct_ocr(text)
+    summary = summarize_text(corrected, lang)
 
-        return text, summary[0]["summary_text"]
+    file_path = save_summary(summary)
+    return corrected, summary, file_path
 
-    except Exception as e:
-        return f"Erreur : {e}", "Erreur."
+# -------- Télécharger --------
+def save_summary(summary):
+    tmp = tempfile.NamedTemporaryFile(delete=False, suffix=".txt")
+    tmp.write(summary.encode("utf-8"))
+    tmp.close()
+    return tmp.name
 
-# ---------- INTERFACE ----------
-with gr.Blocks(title="OCR & Résumé (Image + Fichier)") as demo:
+# -------- INTERFACE --------
+with gr.Blocks(title="OCR & Résumé Intelligent") as demo:
     gr.Markdown("""
-    # 🧠 OCR & Résumé intelligent
-    - 🖼️ Images : OCR + résumé
-    - 📄 Fichiers : TXT / DOCX / PPTX
+    # 🧠 OCR & Résumé Intelligent
+    ✔ Image | TXT | DOCX | PPTX | PDF  
+    ✔ Résumé FR / AR / EN  
+    ✔ Correction OCR  
+    ✔ Téléchargement du résumé
     """)
 
+    lang = gr.Radio(
+        choices=[("Français", "fr"), ("العربية", "ar"), ("English", "en")],
+        value="fr",
+        label="Langue du résumé"
+    )
+
     with gr.Tabs():
 
-        # ----- TAB IMAGE -----
-        with gr.Tab("🖼️ Image (OCR)"):
-            img_input = gr.Image(type="pil", label="Téléverser une image")
-            img_btn = gr.Button("Extraire & Résumer", variant="primary")
-            img_text = gr.Textbox(label="Texte extrait", lines=10)
-            img_summary = gr.Textbox(label="Résumé", lines=5)
+        with gr.Tab("🖼️ Image"):
+            img = gr.Image(type="pil")
+            btn_img = gr.Button("Extraire & Résumer", variant="primary")
+            txt_img = gr.Textbox(label="Texte corrigé", lines=10)
+            sum_img = gr.Textbox(label="Résumé", lines=5)
+            dl_img = gr.File(label="📥 Télécharger le résumé")
 
-            img_btn.click(
+            btn_img.click(
                 process_image,
-                inputs=img_input,
-                outputs=[img_text, img_summary]
+                inputs=[img, lang],
+                outputs=[txt_img, sum_img, dl_img]
             )
 
-        # ----- TAB FILE -----
         with gr.Tab("📄 Fichier"):
-            file_input = gr.File(
-                label="Téléverser un fichier",
-                file_types=[".txt", ".docx", ".pptx"]
-            )
-            file_btn = gr.Button("Extraire & Résumer", variant="primary")
-            file_text = gr.Textbox(label="Texte extrait", lines=10)
-            file_summary = gr.Textbox(label="Résumé", lines=5)
+            file = gr.File(file_types=[".txt", ".docx", ".pptx", ".pdf"])
+            btn_file = gr.Button("Extraire & Résumer", variant="primary")
+            txt_file = gr.Textbox(label="Texte corrigé", lines=10)
+            sum_file = gr.Textbox(label="Résumé", lines=5)
+            dl_file = gr.File(label="📥 Télécharger le résumé")
 
-            file_btn.click(
+            btn_file.click(
                 process_file,
-                inputs=file_input,
-                outputs=[file_text, file_summary]
+                inputs=[file, lang],
+                outputs=[txt_file, sum_file, dl_file]
             )
 
 if __name__ == "__main__":