amerlioration ihm
Browse files- .streamlit/config.toml +8 -1
- app/main.py +24 -5
- app/utils.py +34 -0
.streamlit/config.toml
CHANGED
|
@@ -6,4 +6,11 @@ headless = true
|
|
| 6 |
fileWatcherType = "watchdog"
|
| 7 |
|
| 8 |
[browser]
|
| 9 |
-
gatherUsageStats = false
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
fileWatcherType = "watchdog"
|
| 7 |
|
| 8 |
[browser]
|
| 9 |
+
gatherUsageStats = false
|
| 10 |
+
|
| 11 |
+
[theme]
|
| 12 |
+
base="dark"
|
| 13 |
+
primaryColor="#4CAF50"
|
| 14 |
+
backgroundColor="#0E1117"
|
| 15 |
+
secondaryBackgroundColor="#262730"
|
| 16 |
+
textColor="#FAFAFA"
|
app/main.py
CHANGED
|
@@ -10,7 +10,7 @@ import logging
|
|
| 10 |
|
| 11 |
|
| 12 |
confidence_threshold=0.4
|
| 13 |
-
entropy_threshold=
|
| 14 |
|
| 15 |
logging.basicConfig(
|
| 16 |
level=logging.DEBUG, # ou logging.DEBUG
|
|
@@ -356,10 +356,29 @@ if page == pages[4]:
|
|
| 356 |
st.image(res['image_obj'], caption=f"Image : {res['file_name']}", use_container_width=True)
|
| 357 |
st.markdown("### 🔍 Résultat de la prédiction")
|
| 358 |
st.write(f"📁 Nom du fichier : `{res['file_name']}`")
|
| 359 |
-
|
| 360 |
-
|
| 361 |
-
|
| 362 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 363 |
if is_online:
|
| 364 |
if "rapide" in mode_api:
|
| 365 |
st.info(f"🧠 **{mode_api}** — 1 modèle utilisé : `{res['models_names'][0]}`")
|
|
|
|
| 10 |
|
| 11 |
|
| 12 |
confidence_threshold=0.4
|
| 13 |
+
entropy_threshold=2
|
| 14 |
|
| 15 |
logging.basicConfig(
|
| 16 |
level=logging.DEBUG, # ou logging.DEBUG
|
|
|
|
| 356 |
st.image(res['image_obj'], caption=f"Image : {res['file_name']}", use_container_width=True)
|
| 357 |
st.markdown("### 🔍 Résultat de la prédiction")
|
| 358 |
st.write(f"📁 Nom du fichier : `{res['file_name']}`")
|
| 359 |
+
|
| 360 |
+
predicted_classes = res['models_predictions']
|
| 361 |
+
unique_classes = set(predicted_classes)
|
| 362 |
+
if len(unique_classes) == 1 or not is_online:
|
| 363 |
+
pred_plant, pred_disease = res['pred_class_name'].split('_', 1)
|
| 364 |
+
st.success(f"🌱 **Plante predite :** `{pred_plant}`")
|
| 365 |
+
st.success(f"🩺 **Maladie predite :** `{pred_disease}`")
|
| 366 |
+
if len(predicted_classes)>1:
|
| 367 |
+
st.info("Prédiction fiable. Les modèles sont unanimes.")
|
| 368 |
+
|
| 369 |
+
true_plant, true_disease = extract_class_from_filename(res['file_name'])
|
| 370 |
+
if true_plant is not None and true_disease is not None:
|
| 371 |
+
with st.expander("🔍 **Vérité terrain :**"):
|
| 372 |
+
st.write(f"🌱 Plante réelle : `{true_plant}`")
|
| 373 |
+
st.write(f"🩺 Maladie réelle : `{true_disease}`")
|
| 374 |
+
else:
|
| 375 |
+
st.info("ℹ️ Vérité terrain non disponible pour cette image.")
|
| 376 |
+
|
| 377 |
+
st.write(f"📊 Confiance : **{res['confidence']*100:.2f}%**")
|
| 378 |
+
st.write(f"📈 Entropie : **{res['entropy']:.3f}**")
|
| 379 |
+
else:
|
| 380 |
+
st.warning(f"⚠️ Prédiction incertaine : `{res['pred_class_name']}`")
|
| 381 |
+
st.error("Les modèles ne sont pas d'accord. Veuillez vérifier l'image sur le terrain ou consulter un expert avant de décider d'un traitement.")
|
| 382 |
if is_online:
|
| 383 |
if "rapide" in mode_api:
|
| 384 |
st.info(f"🧠 **{mode_api}** — 1 modèle utilisé : `{res['models_names'][0]}`")
|
app/utils.py
CHANGED
|
@@ -92,6 +92,11 @@ class_names = [
|
|
| 92 |
from typing import List, Any
|
| 93 |
from collections import namedtuple
|
| 94 |
from dataclasses import dataclass
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 95 |
@dataclass
|
| 96 |
class PredictionResult:
|
| 97 |
pred_class_name: str
|
|
@@ -106,6 +111,35 @@ class PredictionResult:
|
|
| 106 |
models_uncertainties: List[bool]
|
| 107 |
gradcam_images: List[Any]
|
| 108 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 109 |
|
| 110 |
def predict_via_api(image_pil, api_url, mode="single",show_heatmap=False,default_model="EfficientNetV2M"):
|
| 111 |
logging.info("🖼️ Préparation de l'image pour l'envoi à l'API...")
|
|
|
|
| 92 |
from typing import List, Any
|
| 93 |
from collections import namedtuple
|
| 94 |
from dataclasses import dataclass
|
| 95 |
+
|
| 96 |
+
|
| 97 |
+
|
| 98 |
+
|
| 99 |
+
|
| 100 |
@dataclass
|
| 101 |
class PredictionResult:
|
| 102 |
pred_class_name: str
|
|
|
|
| 111 |
models_uncertainties: List[bool]
|
| 112 |
gradcam_images: List[Any]
|
| 113 |
|
| 114 |
+
import re
|
| 115 |
+
import os
|
| 116 |
+
|
| 117 |
+
def extract_class_from_filename(filename):
|
| 118 |
+
"""
|
| 119 |
+
Utilise une expression régulière pour extraire 'Plante - Maladie' depuis le nom de fichier.
|
| 120 |
+
Exemple de formats acceptés :
|
| 121 |
+
Rice___Blast___BLAST1_017.jpg
|
| 122 |
+
Blueberry___Healthy___0bfbcb95-30cb-487c-b9b9-49e4edd4af9f___RS_HL 0534.JPG
|
| 123 |
+
Peach___Bacterial_spot___0a620811-30ba-42bf-ace2-8888d4db6dcd___Rutg._Bact.S 1782.JPG
|
| 124 |
+
"""
|
| 125 |
+
try:
|
| 126 |
+
base_name = os.path.basename(filename)
|
| 127 |
+
name_without_ext = os.path.splitext(base_name)[0]
|
| 128 |
+
|
| 129 |
+
# Regex : on cherche une structure 'texte - texte'
|
| 130 |
+
match = re.search(r'^([A-Za-z]+)_+([A-Za-z_]+)', name_without_ext)
|
| 131 |
+
|
| 132 |
+
if match:
|
| 133 |
+
plant = match.group(1).strip()
|
| 134 |
+
disease = match.group(2).strip()
|
| 135 |
+
return plant, disease
|
| 136 |
+
else:
|
| 137 |
+
return None, None
|
| 138 |
+
except Exception as e:
|
| 139 |
+
print(f"Erreur d'extraction : {e}")
|
| 140 |
+
return None, None
|
| 141 |
+
|
| 142 |
+
|
| 143 |
|
| 144 |
def predict_via_api(image_pil, api_url, mode="single",show_heatmap=False,default_model="EfficientNetV2M"):
|
| 145 |
logging.info("🖼️ Préparation de l'image pour l'envoi à l'API...")
|