Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
|
@@ -4,12 +4,11 @@ import unicodedata
|
|
| 4 |
from paddleocr import PaddleOCR
|
| 5 |
|
| 6 |
# =================================================
|
| 7 |
-
# OCR Paddle
|
| 8 |
# =================================================
|
| 9 |
ocr = PaddleOCR(
|
| 10 |
lang="fr",
|
| 11 |
-
use_textline_orientation=True
|
| 12 |
-
show_log=False
|
| 13 |
)
|
| 14 |
|
| 15 |
# =================================================
|
|
@@ -23,7 +22,6 @@ def normalize(text: str) -> str:
|
|
| 23 |
|
| 24 |
# =================================================
|
| 25 |
# Titres possibles de la colonne 2
|
| 26 |
-
# (casse ignorée automatiquement)
|
| 27 |
# =================================================
|
| 28 |
COL_TITLES = {
|
| 29 |
"designation",
|
|
@@ -33,7 +31,7 @@ COL_TITLES = {
|
|
| 33 |
}
|
| 34 |
|
| 35 |
# =================================================
|
| 36 |
-
# Mots à ignorer
|
| 37 |
# =================================================
|
| 38 |
IGNORE_KEYWORDS = {
|
| 39 |
"prix", "ht", "htva", "tva", "ttc",
|
|
@@ -50,14 +48,14 @@ def extract_second_column(image):
|
|
| 50 |
img = np.array(image)
|
| 51 |
|
| 52 |
# -------------------------------------------------
|
| 53 |
-
#
|
| 54 |
# -------------------------------------------------
|
| 55 |
h, w = img.shape[:2]
|
| 56 |
if w > h:
|
| 57 |
img = np.rot90(img, 1)
|
| 58 |
|
| 59 |
# -------------------------------------------------
|
| 60 |
-
#
|
| 61 |
# -------------------------------------------------
|
| 62 |
result = ocr.predict(img)
|
| 63 |
if not result:
|
|
@@ -82,7 +80,7 @@ def extract_second_column(image):
|
|
| 82 |
return "Pas assez de texte exploitable."
|
| 83 |
|
| 84 |
# -------------------------------------------------
|
| 85 |
-
#
|
| 86 |
# -------------------------------------------------
|
| 87 |
col_x = None
|
| 88 |
for text, x, y in blocks:
|
|
@@ -98,7 +96,7 @@ def extract_second_column(image):
|
|
| 98 |
return "Titre de la colonne cible non détecté."
|
| 99 |
|
| 100 |
# -------------------------------------------------
|
| 101 |
-
#
|
| 102 |
# -------------------------------------------------
|
| 103 |
X_THRESHOLD = 55
|
| 104 |
column_blocks = [
|
|
@@ -110,12 +108,12 @@ def extract_second_column(image):
|
|
| 110 |
return "Colonne détectée mais vide."
|
| 111 |
|
| 112 |
# -------------------------------------------------
|
| 113 |
-
#
|
| 114 |
# -------------------------------------------------
|
| 115 |
column_blocks.sort(key=lambda e: e[2])
|
| 116 |
|
| 117 |
# -------------------------------------------------
|
| 118 |
-
#
|
| 119 |
# -------------------------------------------------
|
| 120 |
merged = []
|
| 121 |
current = ""
|
|
@@ -141,25 +139,18 @@ def extract_second_column(image):
|
|
| 141 |
merged.append(current.strip())
|
| 142 |
|
| 143 |
# -------------------------------------------------
|
| 144 |
-
#
|
| 145 |
# -------------------------------------------------
|
| 146 |
final = []
|
| 147 |
for line in merged:
|
| 148 |
nt = normalize(line)
|
| 149 |
|
| 150 |
-
# ignorer le titre de colonne
|
| 151 |
if nt in COL_TITLES:
|
| 152 |
continue
|
| 153 |
-
|
| 154 |
-
# longueur minimale
|
| 155 |
if len(nt) < 5:
|
| 156 |
continue
|
| 157 |
-
|
| 158 |
-
# ignorer lignes trop numériques
|
| 159 |
if sum(c.isdigit() for c in line) > len(line) / 3:
|
| 160 |
continue
|
| 161 |
-
|
| 162 |
-
# règle métier : commence par majuscule
|
| 163 |
if not line[0].isupper():
|
| 164 |
continue
|
| 165 |
|
|
@@ -168,13 +159,10 @@ def extract_second_column(image):
|
|
| 168 |
if not final:
|
| 169 |
return "Aucune cellule texte valide trouvée."
|
| 170 |
|
| 171 |
-
# -------------------------------------------------
|
| 172 |
-
# 7. Résultat numéroté
|
| 173 |
-
# -------------------------------------------------
|
| 174 |
return "\n".join(f"{i+1}. {line}" for i, line in enumerate(final))
|
| 175 |
|
| 176 |
# =================================================
|
| 177 |
-
# Interface Gradio
|
| 178 |
# =================================================
|
| 179 |
demo = gr.Interface(
|
| 180 |
fn=extract_second_column,
|
|
@@ -182,9 +170,8 @@ demo = gr.Interface(
|
|
| 182 |
outputs=gr.Textbox(label="Contenu de la colonne 2"),
|
| 183 |
title="Extraction fiable de la colonne 2",
|
| 184 |
description=(
|
| 185 |
-
"Extraction
|
| 186 |
-
"(Désignation, DESIGNATIONS, Description, Description des services)
|
| 187 |
-
"à partir de tableaux scannés."
|
| 188 |
)
|
| 189 |
)
|
| 190 |
|
|
|
|
| 4 |
from paddleocr import PaddleOCR
|
| 5 |
|
| 6 |
# =================================================
|
| 7 |
+
# OCR Paddle (HF compatible)
|
| 8 |
# =================================================
|
| 9 |
ocr = PaddleOCR(
|
| 10 |
lang="fr",
|
| 11 |
+
use_textline_orientation=True
|
|
|
|
| 12 |
)
|
| 13 |
|
| 14 |
# =================================================
|
|
|
|
| 22 |
|
| 23 |
# =================================================
|
| 24 |
# Titres possibles de la colonne 2
|
|
|
|
| 25 |
# =================================================
|
| 26 |
COL_TITLES = {
|
| 27 |
"designation",
|
|
|
|
| 31 |
}
|
| 32 |
|
| 33 |
# =================================================
|
| 34 |
+
# Mots à ignorer
|
| 35 |
# =================================================
|
| 36 |
IGNORE_KEYWORDS = {
|
| 37 |
"prix", "ht", "htva", "tva", "ttc",
|
|
|
|
| 48 |
img = np.array(image)
|
| 49 |
|
| 50 |
# -------------------------------------------------
|
| 51 |
+
# Rotation automatique si image couchée
|
| 52 |
# -------------------------------------------------
|
| 53 |
h, w = img.shape[:2]
|
| 54 |
if w > h:
|
| 55 |
img = np.rot90(img, 1)
|
| 56 |
|
| 57 |
# -------------------------------------------------
|
| 58 |
+
# OCR
|
| 59 |
# -------------------------------------------------
|
| 60 |
result = ocr.predict(img)
|
| 61 |
if not result:
|
|
|
|
| 80 |
return "Pas assez de texte exploitable."
|
| 81 |
|
| 82 |
# -------------------------------------------------
|
| 83 |
+
# Détection du X de la colonne 2 via le titre
|
| 84 |
# -------------------------------------------------
|
| 85 |
col_x = None
|
| 86 |
for text, x, y in blocks:
|
|
|
|
| 96 |
return "Titre de la colonne cible non détecté."
|
| 97 |
|
| 98 |
# -------------------------------------------------
|
| 99 |
+
# Sélection des blocs de la colonne
|
| 100 |
# -------------------------------------------------
|
| 101 |
X_THRESHOLD = 55
|
| 102 |
column_blocks = [
|
|
|
|
| 108 |
return "Colonne détectée mais vide."
|
| 109 |
|
| 110 |
# -------------------------------------------------
|
| 111 |
+
# Tri vertical
|
| 112 |
# -------------------------------------------------
|
| 113 |
column_blocks.sort(key=lambda e: e[2])
|
| 114 |
|
| 115 |
# -------------------------------------------------
|
| 116 |
+
# Fusion intelligente des lignes OCR
|
| 117 |
# -------------------------------------------------
|
| 118 |
merged = []
|
| 119 |
current = ""
|
|
|
|
| 139 |
merged.append(current.strip())
|
| 140 |
|
| 141 |
# -------------------------------------------------
|
| 142 |
+
# Nettoyage final
|
| 143 |
# -------------------------------------------------
|
| 144 |
final = []
|
| 145 |
for line in merged:
|
| 146 |
nt = normalize(line)
|
| 147 |
|
|
|
|
| 148 |
if nt in COL_TITLES:
|
| 149 |
continue
|
|
|
|
|
|
|
| 150 |
if len(nt) < 5:
|
| 151 |
continue
|
|
|
|
|
|
|
| 152 |
if sum(c.isdigit() for c in line) > len(line) / 3:
|
| 153 |
continue
|
|
|
|
|
|
|
| 154 |
if not line[0].isupper():
|
| 155 |
continue
|
| 156 |
|
|
|
|
| 159 |
if not final:
|
| 160 |
return "Aucune cellule texte valide trouvée."
|
| 161 |
|
|
|
|
|
|
|
|
|
|
| 162 |
return "\n".join(f"{i+1}. {line}" for i, line in enumerate(final))
|
| 163 |
|
| 164 |
# =================================================
|
| 165 |
+
# Interface Gradio
|
| 166 |
# =================================================
|
| 167 |
demo = gr.Interface(
|
| 168 |
fn=extract_second_column,
|
|
|
|
| 170 |
outputs=gr.Textbox(label="Contenu de la colonne 2"),
|
| 171 |
title="Extraction fiable de la colonne 2",
|
| 172 |
description=(
|
| 173 |
+
"Extraction adaptative de la deuxième colonne "
|
| 174 |
+
"(Désignation, DESIGNATIONS, Description, Description des services)."
|
|
|
|
| 175 |
)
|
| 176 |
)
|
| 177 |
|