Add application file

.gitattributes

@@ -29,3 +29,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zstandard filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text

app.py

@@ -0,0 +1,122 @@
+import streamlit as st
+import requests
+from PIL import Image
+import json
+import numpy as np
+from tensorflow.keras import models
+import cv2
+from tensorflow.keras.preprocessing.image import array_to_img
+import glob
+import os
+
+
+
+ex_path = ""
+images_paths = [image_name for image_name in glob.glob(os.path.join(ex_path, "image/*"))]
+dico_paths = {}
+dico_paths['  '] = None
+for img in images_paths:
+    dico_paths[img[-36:-23]] = img
+    
+def get_true_msk(img_path):
+    msk_name = img_path[-36:-15]+'gtFine_labelIds.png'
+    return [msk_link for msk_link in glob.glob(os.path.join(ex_path, "mask\\", msk_name))] 
+    
+def dice_coeff(y_true, y_pred):
+    smooth = 1.
+    y_true_f = K.flatten(y_true)
+    y_pred_f = K.flatten(y_pred)
+    intersection = K.sum(y_true_f * y_pred_f)
+    score = (2. * intersection + smooth) / (K.sum(y_true_f) + K.sum(y_pred_f) + smooth)
+    return score
+    
+def recolor(img):
+    prediction_color = {0:[0, 0, 0],          
+                        1:[204, 204, 204],       
+                        2:[148, 148, 148],        
+                        3:[220, 220, 0],     
+                        4:[111, 196, 53],      
+                        5:[188, 230, 254],      
+                        6:[191, 30, 62],      
+                        7:[3, 50, 126]}        
+    new_img = np.zeros((*img.shape, 3))
+    for k, v in prediction_color.items():
+        new_img[img==k] = v
+    new_img = new_img.astype('uint8')
+    return new_img 
+
+
+def encode_labels(mask):
+    mapping_8 = {
+        0:0, 1:0, 2:0, 3:0, 4:0, 5:0, 6:0,
+        7:1, 8:1, 9:1, 10:1,
+        11:2, 12:2, 13:2, 14:2, 15:2, 16:2,
+        17:3, 18:3, 19:3, 20:3,  
+        21:4, 22:4,
+        23:5,
+        24:6, 25:6,
+        26:7, 27:7, 28:7, 29:7, 30:7, 31:7, 32:7, 33:7, -1:7
+            }
+    label_mask = np.zeros_like(mask)
+    for k in mapping_8:
+        label_mask[mask == k] = mapping_8[k]
+    return label_mask    
+    
+def request_prediction(option):
+
+    #Model & File
+    model = models.load_model('model_best.h5', custom_objects={"dice_coeff": dice_coeff})
+    file_in = open(option ,'rb')
+    
+    #Bloc preprocessing    
+    y_img = Image.open(file_in)
+    y_img = np.array(y_img)
+    y_img = cv2.resize(y_img, (128, 128))
+    
+    #Bloc prédiction
+    batch_img = np.empty((32, 128, 128, 3))
+    batch_img[0] = y_img
+    y_pred = model.predict(batch_img)
+
+    #Bloc image
+    printable_pred = np.argmax(y_pred[0], axis=2)
+    printable_pred = recolor(printable_pred)
+    img = array_to_img(printable_pred)
+    
+    return img
+    
+
+def main():
+
+
+
+    st.title('Mask Prediction')
+    option = st.selectbox('Choisissez une image',(dico_paths.keys()))
+    st.write('You selected:', option)    
+
+
+    predict_btn = st.button('Prédire')
+    if predict_btn:
+        pred = None
+        pred = request_prediction(dico_paths[option])
+
+        file_img = open(dico_paths[option] ,'rb')
+        img = Image.open(file_img)
+        
+        file_msk = open(get_true_msk(dico_paths[option])[0] ,'rb')
+        msk = Image.open(file_msk)
+        
+        msk = np.asarray(msk, dtype="float32")
+        msk = encode_labels(msk)
+        msk = cv2.resize(msk, (128, 128))
+        msk = np.asarray(msk, dtype="int32")
+        msk = array_to_img(recolor(msk))
+        
+        st.image(img, caption = 'Origine', channels = 'BGR')
+        st.image(pred, caption = 'Prédiction', width = 600, channels = 'BGR')
+        st.image(msk, caption = 'Label', width = 600, channels = 'BGR')
+
+
+
+if __name__ == '__main__':
+    main()

model_best.h5

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:93d10b239e8db5fdc2f9cfcc3eec79f58da7b4b679815379f3c513376153016c
+size 414763072

requirements.txt

@@ -0,0 +1,5 @@
+numpy==1.20.3
+opencv_python==4.5.5.64
+Pillow==9.2.0
+streamlit==1.10.0
+tensorflow==2.8.0