Merge branch 'main' into 'victor_prgm'

# Conflicts:
#   supervised_learning/create_classifiers.py
#   supervised_learning/detection.py

data/val/predicted_labels/0839.csv

+0,48,127,193,fvert,0.9488638965462507
+608,144,735,289,frouge,0.9108585220743701
+272,176,399,321,fvert,0.8895836594445553
+400,128,527,273,fvert,0.877665580856861
+560,416,687,561,frouge,0.822439042674998
+320,336,447,481,frouge,0.7707272520675015
+128,32,255,177,fvert,0.7420041200043779

data/val/predicted_labels/0844.csv

+264,168,454,385,stop,0.9967170965308115
+240,656,367,801,fvert,0.996328291799753
+416,48,543,193,frouge,0.9737347317354985
+224,400,351,545,fvert,0.9724734390986917
+0,96,190,313,forange,0.9532640691139742
+112,400,239,545,fvert,0.9476349394512237
+480,456,670,673,frouge,0.9380977779581677
+400,352,527,497,forange,0.9218452514529774
+608,592,735,737,fvert,0.9048974297689133
+0,304,127,449,fvert,0.9015444544626695
+54,378,482,867,fvert,0.8505643526126447
+160,208,287,353,frouge,0.847427848014558
+336,448,463,593,forange,0.7811306591945232
+144,512,271,657,fvert,0.7220600566071843
+480,168,670,385,fvert,0.7199181737080304
+608,336,735,481,fvert,0.7186611921927715
+160,768,287,913,fvert,0.7015408781850739

data/val/predicted_labels/0846.csv

+256,656,383,801,frouge,0.9999998649101208
+608,192,735,337,ceder,0.999985728904932
+368,336,495,481,frouge,0.9951889989104354
+162,486,590,975,frouge,0.9928645239007842
+368,624,495,769,frouge,0.9874774247331431
+360,48,550,265,fvert,0.9869204915415256
+368,752,495,897,frouge,0.9849570651307225
+496,848,623,993,interdiction,0.9837698803541571
+560,448,687,593,ceder,0.9738150434245632
+512,272,639,417,danger,0.9716125769371682
+208,0,335,145,ceder,0.9695208218905375
+608,0,735,145,danger,0.956130554303084
+224,256,351,401,obligation,0.9559719636347589
+160,704,287,849,frouge,0.9243023917482435
+96,496,223,641,interdiction,0.9208845929912137
+64,624,191,769,danger,0.9164699794834144
+216,384,406,601,frouge,0.9014073102428324
+288,832,415,977,frouge,0.8952890587426112
+144,848,271,993,interdiction,0.883629789242684
+112,176,239,321,frouge,0.8823683973255505
+432,464,559,609,frouge,0.8437691194308651
+400,224,527,369,fvert,0.7084189612883776

data/val/predicted_labels/0852.csv

+432,48,622,265,forange,0.9544132668619814
+448,304,575,449,interdiction,0.9533305453674042
+96,736,223,881,frouge,0.9296377777612068
+624,656,751,801,frouge,0.8873901292257897
+368,448,495,593,forange,0.8668555533944166
+304,848,431,993,interdiction,0.8532575803217125
+144,304,271,449,forange,0.8458316138406432
+624,464,751,609,forange,0.8367287794984228
+208,480,335,625,frouge,0.7821556570474244
+64,112,191,257,frouge,0.781953628921386
+360,696,550,913,interdiction,0.7303241781141279

data/val/predicted_labels/0864.csv

+384,96,574,313,stop,0.9999961786208349
+464,272,591,417,frouge,0.961581906672076
+312,336,502,553,fvert,0.8588571631373797

data/val/predicted_labels/0871.csv

+464,464,591,609,danger,0.9879619342232258
+464,64,591,209,danger,0.9857065269685417
+396,612,681,938,interdiction,0.9824499929161948
+464,336,591,481,danger,0.9667308314787896
+528,208,655,353,danger,0.9369233084512567
+416,192,543,337,danger,0.8380811438105122
+324,216,752,705,danger,0.7609197419181045
+272,848,399,993,frouge,0.7601510281044455
+672,704,799,849,danger,0.7237105096500488

data/val/predicted_labels/0878.csv

+160,496,287,641,frouge,0.9819131673246628
+576,64,703,209,forange,0.9562738506195138
+16,672,143,817,fvert,0.9101673803587994
+112,64,239,209,forange,0.9006204916596843
+160,672,287,817,forange,0.8538708189587332
+456,408,646,625,frouge,0.8525610810660433
+400,576,527,721,fvert,0.7571704612873768
+576,560,703,705,fvert,0.7314249488657094
+0,96,127,241,frouge,0.7190222044228405

supervised_learning/config.py

@@ -4,6 +4,6 @@ TRAINING_IMAGE_FOLDER_PATH = "./data/train/images/"
 TRAINING_LABEL_FOLDER_PATH = "./data/train/labels/"
 VAL_IMAGE_FOLDER_PATH = "./data/val/images/"
 VAL_LABEL_FOLDER_PATH = "./data/val/labels/"
-PREDICTION_LABEL_FOLDER_PATH = "./data/train/predicted_labels/"
+PREDICTION_LABEL_FOLDER_PATH = "./data/val/predicted_labels/"
 CLASSIFIERS_FOLDER_PATH = "./supervised_learning/classifiers/saves/"
 STEP_SIZE = 16
\ No newline at end of file

supervised_learning/create_classifiers.py

@@ -99,6 +99,7 @@ for classe in CLASSES:
     if classe not in ['ff', 'empty']:
         datasets["train"][classe]["X"], datasets["train"][classe]["Y"] = create_binary_classification_dataset(datas_train, classe)
         datasets["val"][classe]["X"], datasets["val"][classe]["Y"] = create_binary_classification_dataset(datas_val, classe)
+        
 
 # Create a dataset for general light classifier
 datasets["train"]["feux"]["X"], datasets["train"]["feux"]["Y"] = create_binary_classification_dataset(datas_train, "feux")
@@ -135,7 +136,6 @@ for classe in classifiers.keys():
         y_pred = classifiers[classe].predict(X_val)
         print(f"Précision pour panneaux {classe}: {np.mean(y_pred == y_val)}")
 
-
 # ------------- SAVE CLASSIFIERS -----------------
 print("Saving classifiers")
 for classes, model in classifiers.items():

supervised_learning/detection.py

@@ -13,6 +13,7 @@ parser.add_argument('output_file', metavar='DIR', help='Folder of output labels
 # Analyser les arguments de la ligne de commande
 args = parser.parse_args()
 
+
 # Load input folder
 input_folder = args.input_folder
 # Check if input folder exists

supervised_learning/vue.py

-from supervised_learning.detection import *
 from supervised_learning.config import *
 from supervised_learning.utils import *
 from PIL import Image, ImageDraw, ImageFont, ImageColor
@@ -6,8 +5,30 @@ import os
 import csv
 import random
 import matplotlib.pyplot as plt
+import argparse 
+
+# ------------- IMPORT PARSER & ARGS ---------------
+
+parser = argparse.ArgumentParser()
+parser.add_argument('img_folder', metavar='DIR', help='Folder of input images to analyse')
+parser.add_argument('pred_folder', metavar='DIR', help='Folder of predicted labels from the analysis')
+
+# Analyser les arguments de la ligne de commande
+args = parser.parse_args()
+
+
+# Load input folder
+img_folder = args.img_folder
+# Check if input folder exists
+if not os.path.exists(img_folder):
+    raise FileNotFoundError(f"[Error] Folder '{img_folder}' does not exist.")
+
+# Load input folder
+pred_folder = args.pred_folder
+# Check if input folder exists
+if not os.path.exists(pred_folder):
+    raise FileNotFoundError(f"[Error] Folder '{pred_folder}' does not exist.")
 
-#detecteur(VAL_IMAGE_FOLDER_PATH, PREDICTION_LABEL_FOLDER_PATH)
 
 def calculate_text_size(text, font):
     # calculate text size based on font properties
@@ -80,9 +101,9 @@ fig, axs = plt.subplots(2, 5, figsize=(12, 6))
 for ax in axs.ravel():
     #images_dir = os.path.join("train", "images")
     #labels_dir = os.path.join("train", "labels")
-    image_name = random.choice(os.listdir(VAL_IMAGE_FOLDER_PATH))
-    image_path = os.path.join(VAL_IMAGE_FOLDER_PATH, image_name)
-    csv_path = os.path.join(PREDICTION_LABEL_FOLDER_PATH, image_name[:-4] + ".csv")   
+    image_name = random.choice(os.listdir(img_folder))
+    image_path = os.path.join(img_folder, image_name)
+    csv_path = os.path.join(pred_folder, image_name[:-4] + ".csv")   
     # Call visualize_image function to modify the image
     image_to_display = visualize_image(image_path, csv_path)
     

utils/selective_search.py

+import cv2
+import matplotlib.pyplot as plt
+
+def selective_search(image, visualize=True, visulize_count=100):
+    # Convert image to BGR format for OpenCV
+    image_bgr = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
+    
+    # Initialiser la recherche sélective
+    ss = cv2.ximgproc.segmentation.createSelectiveSearchSegmentation()
+    ss.setBaseImage(image_bgr)
+
+    # Utiliser la recherche sélective en mode rapide (ou en mode qualité)
+    ss.switchToSelectiveSearchFast()  # Pour la recherche rapide
+    # ss.switchToSelectiveSearchQuality()  # Pour une recherche plus précise
+
+    # Obtenir les régions candidates
+    roi = ss.process()
+
+    if visualize:
+        # Dessiner les régions candidates sur l'image
+        for (x, y, w, h) in roi[:visulize_count]:  # Limiter à 100 régions pour la visualisation
+            cv2.rectangle(image_bgr, (x, y), (x + w, y + h), (0, 255, 0), 2)
+
+        # Convertir l'image BGR en RGB pour Matplotlib
+        image_rgb = cv2.cvtColor(image_bgr, cv2.COLOR_BGR2RGB)
+        
+        # Afficher l'image avec les régions candidates
+        plt.figure(figsize=(10, 10))
+        plt.imshow(image_rgb)
+        plt.axis('off')  # Cacher les axes pour une meilleure visualisation
+        plt.show()
+        
+    return roi
+
+img = cv2.imread("./data/val/images/0846.jpg")
+img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)  # Assurer que l'image est en RGB
+result = selective_search(img_rgb)