Chest_X-Ray_Covid_Detection_DL / custom_functions.py
ChaoukiBenzekri's picture
Add costom_functions
2ec1466
raw
history blame
17.7 kB
import streamlit as st
import uuid
import re
import numpy as np
import cv2
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
import plotly.graph_objects as go
import plotly.figure_factory as ff
import tensorflow as tf
import keras
from keras.layers import Conv2D
from tensorflow.keras.preprocessing.image import img_to_array, array_to_img
from keras.models import load_model
# Afficher le nom et l'icone des réseaux d'un membre
def show_profile(name, linkedin_url, github_url):
st.markdown("""
<style>
.icon-img {
height: 20px; # Taille des icônes
width: 20px;
margin-top: 0px;
margin-left: 5px;
margin-right:5px;
}
.name-with-icons {
display: flex-shrink;
align-items: center;
margin-top: -5px;
margin-bottom: 0px;
}
.name-with-icons:first-child {
margin-top: -5px; /* Marge négative pour le premier élément */
}
</style>
""", unsafe_allow_html = True)
st.markdown(f"""
<div class='name-with-icons'>
<a href='{github_url}' target='_blank'><img class='icon-img' src='https://upload.wikimedia.org/wikipedia/commons/9/91/Octicons-mark-github.svg'></a>
<a href='{linkedin_url}' target='_blank'><img class='icon-img' src='https://upload.wikimedia.org/wikipedia/commons/c/ca/LinkedIn_logo_initials.png'></a>
{name}
</div>
""", unsafe_allow_html = True)
# Création de bloc de texte esthétiques
def create_styled_box(text, text_color, background_color, alignment = 'left', display = 'block'):
unique_id = uuid.uuid4().hex
# Ajout de styles CSS pour le cadre avec une couleur de fond et une couleur de texte personnalisée
st.markdown(f"""
<style>
.styled-box-{unique_id} {{
background-color: {background_color}; /* couleur de fond */
color: {text_color}; /* couleur de texte */
padding: 10px 20px; /* espace intérieur vertical et horizontal */
border: 1px solid {text_color}; /* bordure de la même couleur que le texte */
border-radius: 8px; /* bord arrondi */
font-size: 16px; /* taille de la police */
font-weight: regular; /* format de la police */
box-shadow: 0 4px 8px rgba(0,0,0,0.2); /* ombre pour donner de la profondeur */
text-align: {alignment}; /* alignement du texte */
display: {display};
}}
</style>
""", unsafe_allow_html = True)
# Afficher le texte dans le cadre stylisé
st.markdown(f"""
<div class="styled-box-{unique_id}">{text}</div>
""", unsafe_allow_html = True)
# Fonction pour calculer l'intensité lumineuse moyenne d'une image
def calc_mean_intensity(image_path):
img = mpimg.imread(image_path)
# Convertir en nuances de gris si l'image est en couleur
if img.ndim == 3:
img_gray = np.dot(img[...,:3], [0.2989, 0.5870, 0.1140])
else:
img_gray = img
return np.mean(img_gray)
# Fonction pour extraire les sources depuis les urls
def source_extract(url):
pattern = re.compile(r'https?://(?:www\.)?([^/]+)')
match = pattern.search(url)
if match:
return match.group(1)
else:
return None
# Fonctions de plot des métriques
def plot_loss_curve(history):
fig = go.Figure()
fig.add_trace(go.Scatter(x = list(range(len(history['loss']))),
y = history['loss'],
mode = 'lines+markers',
name = 'Perte d\'entraînement',
marker = dict(color = 'lightblue')))
fig.add_trace(go.Scatter(x = list(range(len(history['val_loss']))),
y = history['val_loss'],
mode = 'lines+markers',
name = 'Perte de validation',
marker = dict(color = 'salmon')))
fig.update_layout(title = dict(text = "Courbe de Perte", font = dict(color = 'white')),
xaxis_title = dict(text = "Époque", font = dict(color = 'white')),
yaxis_title = dict(text = "Perte", font = dict(color = 'white')),
template = 'plotly_white',
paper_bgcolor = 'rgba(0,0,0,0)',
plot_bgcolor = 'rgba(0,0,0,0)',
legend = dict(font = dict(color = 'white')))
st.plotly_chart(fig)
def plot_precision_curve(history):
fig = go.Figure()
fig.add_trace(go.Scatter(x = list(range(len(history['precision']))),
y = history['precision'],
mode = 'lines+markers',
name = "Précision d'entraînement",
marker = dict(color = 'lightblue')))
fig.add_trace(go.Scatter(x = list(range(len(history['val_precision']))),
y = history['val_precision'],
mode = 'lines+markers',
name = 'Précision de validation',
marker = dict(color = 'salmon')))
fig.update_layout(title = dict(text = "Courbe de Précision", font = dict(color = 'white')),
xaxis_title=dict(text = "Époque", font = dict(color = 'white')),
yaxis_title=dict(text = "Précision", font = dict(color = 'white')),
template = 'plotly_white',
paper_bgcolor = 'rgba(0,0,0,0)',
plot_bgcolor = 'rgba(0,0,0,0)',
legend = dict(font = dict(color='white')))
st.plotly_chart(fig)
def plot_auc(history):
fig = go.Figure()
fig.add_trace(go.Scatter(x = list(range(len(history['auc']))),
y = history['auc'],
mode = 'lines+markers',
name = "AUC moyen d'entraînement",
marker = dict(color='lightblue')))
fig.add_trace(go.Scatter(x = list(range(len(history['val_auc']))),
y = history['val_auc'],
mode = 'lines+markers',
name = 'AUC moyen de validation',
marker=dict(color = 'salmon')))
fig.update_layout(title = "Courbe de AUC-ROC",
xaxis_title = "Époque",
yaxis_title = "Area Under Curve",
template = 'plotly_white',
paper_bgcolor = 'rgba(0,0,0,0)',
plot_bgcolor = 'rgba(0,0,0,0)',
legend = dict(font = dict(color = 'white')),
xaxis = dict(tickfont = dict(color = 'white')),
yaxis = dict(tickfont = dict(color = 'white')),
title_font = dict(color = 'white'))
st.plotly_chart(fig)
def plot_f1_score(history):
fig = go.Figure()
fig.add_trace(go.Scatter(x = list(range(len(history['f1_score']))),
y=np.mean(history['f1_score'], axis = 1),
mode = 'lines+markers',
name = "F1 Score d'entraînement",
marker = dict(color = 'lightblue')))
fig.add_trace(go.Scatter(x = list(range(len(history['val_f1_score']))),
y = np.mean(history['val_f1_score'], axis = 1),
mode = 'lines+markers',
name = 'F1 Score moyen de validation',
marker = dict(color = 'salmon')))
fig.update_layout(title = "Courbe de F1 Score",
xaxis_title = "Époque",
yaxis_title = "F1 Score",
template = 'plotly_white',
paper_bgcolor = 'rgba(0,0,0,0)',
plot_bgcolor = 'rgba(0,0,0,0)',
legend = dict(font = dict(color = 'white')),
xaxis = dict(tickfont = dict(color = 'white')),
yaxis = dict(tickfont = dict(color = 'white')),
title_font = dict(color = 'white'))
st.plotly_chart(fig)
# Plot des matrices de confusion
def plot_CM(matrix):
confusion_matrix = np.array(matrix)
class_names = ['Covid', 'Lung opacity', 'Normal', 'Viral Pneumonia']
st.title('Matrice de Confusion')
fig = ff.create_annotated_heatmap(z = confusion_matrix, x = class_names, y = class_names, colorscale = 'RdBu')
fig.update_layout(
title = 'Matrice de Confusion',
xaxis = dict(title = 'Classe Prédite'),
yaxis = dict(title = 'Classe Réelle')
)
st.plotly_chart(fig)
def plot_CM_ResNetV2():
confusion_lines = [
[192, 3, 7, 2],
[9, 148, 24, 0],
[5, 17, 139, 4],
[1, 0, 4, 165]
]
confusion_matrix = np.array(confusion_lines)
class_names = ['Covid', 'Lung opacity', 'Normal', 'Viral Pneumonia']
size = max(confusion_matrix.shape)
fig = ff.create_annotated_heatmap(z=confusion_matrix, x=class_names, y=class_names, colorscale='Cividis')
fig.update_layout(
width=size*140,
height=size*140,
xaxis=dict(title='Classe Prédite'),
yaxis=dict(title='Classe Réelle')
)
st.plotly_chart(fig)
def plot_CM_ResNet121():
confusion_lines = [
[181, 11, 11, 1],
[8, 148, 25, 0],
[10, 17, 135, 3],
[2, 0, 1, 167]
]
confusion_matrix = np.array(confusion_lines)
class_names = ['Covid', 'Lung opacity', 'Normal', 'Viral Pneumonia']
size = max(confusion_matrix.shape)
fig = ff.create_annotated_heatmap(z=confusion_matrix, x=class_names, y=class_names, colorscale='Cividis')
fig.update_layout(
width=size*140,
height=size*140,
xaxis=dict(title='Classe Prédite'),
yaxis=dict(title='Classe Réelle')
)
st.plotly_chart(fig)
def plot_CM_DenseNet201():
confusion_lines = [
[190, 5, 7, 2],
[6, 156, 19, 0],
[4, 21, 139, 1],
[1, 0, 0, 169]
]
confusion_matrix = np.array(confusion_lines)
class_names = ['Covid', 'Lung opacity', 'Normal', 'Viral Pneumonia']
size = max(confusion_matrix.shape)
fig = ff.create_annotated_heatmap(z=confusion_matrix, x=class_names, y=class_names, colorscale='Cividis')
fig.update_layout(
width=size*140,
height=size*140,
xaxis=dict(title='Classe Prédite'),
yaxis=dict(title='Classe Réelle')
)
st.plotly_chart(fig)
def plot_CM_VGG16():
confusion_lines = [
[178, 5, 9, 2],
[4, 152, 17, 0],
[2, 11, 160, 3],
[0, 0, 4, 175]
]
confusion_matrix = np.array(confusion_lines)
class_names = ['Covid', 'Lung opacity', 'Normal', 'Viral Pneumonia']
size = max(confusion_matrix.shape)
fig = ff.create_annotated_heatmap(z=confusion_matrix, x=class_names, y=class_names, colorscale='Cividis')
fig.update_layout(
width=size*140,
height=size*140,
xaxis=dict(title='Classe Prédite'),
yaxis=dict(title='Classe Réelle')
)
st.plotly_chart(fig)
def plot_CM_VGG19():
confusion_lines = [
[182, 7, 3, 0],
[7, 158, 8, 0],
[8, 21, 142, 5],
[1, 1, 3, 174]
]
confusion_matrix = np.array(confusion_lines)
class_names = ['Covid', 'Lung opacity', 'Normal', 'Viral Pneumonia']
size = max(confusion_matrix.shape)
fig = ff.create_annotated_heatmap(z=confusion_matrix, x=class_names, y=class_names, colorscale='Cividis')
fig.update_layout(
width=size*140,
height=size*140,
xaxis=dict(title='Classe Prédite'),
yaxis=dict(title='Classe Réelle')
)
st.plotly_chart(fig)
def plot_CM_ConvnextTiny():
confusion_lines = [
[122, 11, 19, 0],
[13, 142, 15, 0],
[17, 14, 144, 1],
[4, 1, 7, 168]
]
confusion_matrix = np.array(confusion_lines)
class_names = ['Covid', 'Lung opacity', 'Normal', 'Viral Pneumonia']
size = max(confusion_matrix.shape)
fig = ff.create_annotated_heatmap(z=confusion_matrix, x=class_names, y=class_names, colorscale='Cividis')
fig.update_layout(
width=size*140,
height=size*140,
xaxis=dict(title='Classe Prédite'),
yaxis=dict(title='Classe Réelle')
)
st.plotly_chart(fig)
def plot_CM_ConvnextBase():
confusion_lines = [
[168, 9, 15, 0],
[9, 152, 12, 0],
[12, 8, 153, 3],
[2, 0, 8, 169]
]
confusion_matrix = np.array(confusion_lines)
class_names = ['Covid', 'Lung opacity', 'Normal', 'Viral Pneumonia']
size = max(confusion_matrix.shape)
fig = ff.create_annotated_heatmap(z=confusion_matrix, x=class_names, y=class_names, colorscale='Cividis')
fig.update_layout(
width=size*140,
height=size*140,
xaxis=dict(title='Classe Prédite'),
yaxis=dict(title='Classe Réelle')
)
st.plotly_chart(fig)
def plot_CM_EfficientNet_B4():
confusion_lines = [
[177, 17, 10, 0],
[3, 148, 30, 0],
[1, 13, 151, 0],
[2, 1, 9, 158]
]
confusion_matrix = np.array(confusion_lines)
class_names = ['Covid', 'Lung opacity', 'Normal', 'Viral Pneumonia']
size = max(confusion_matrix.shape)
fig = ff.create_annotated_heatmap(z=confusion_matrix, x=class_names, y=class_names, colorscale='Cividis')
fig.update_layout(
width=size*140,
height=size*140,
xaxis=dict(title='Classe Prédite'),
yaxis=dict(title='Classe Réelle')
)
st.plotly_chart(fig)
def plot_CM_VGG16_FT():
confusion_lines = [
[229, 7, 6, 0],
[1, 198, 16, 1],
[7, 6, 199, 4],
[0, 0, 1, 225]
]
confusion_matrix = np.array(confusion_lines)
class_names = ['Covid', 'Lung opacity', 'Normal', 'Viral Pneumonia']
size = max(confusion_matrix.shape)
fig = ff.create_annotated_heatmap(z=confusion_matrix, x=class_names, y=class_names, colorscale='Cividis')
fig.update_layout(
width=size*140,
height=size*140,
xaxis=dict(title='Classe Prédite'),
yaxis=dict(title='Classe Réelle')
)
st.plotly_chart(fig)
def plot_CM_ResNetFT():
confusion_lines = [
[278, 6, 3, 1],
[3, 242, 16, 0],
[7, 38, 197, 17],
[2, 1, 0, 265]
]
confusion_matrix = np.array(confusion_lines)
class_names = ['Covid', 'Lung opacity', 'Normal', 'Viral Pneumonia']
size = max(confusion_matrix.shape)
fig = ff.create_annotated_heatmap(z=confusion_matrix, x=class_names, y=class_names, colorscale='Cividis')
fig.update_layout(
width=size*140,
height=size*140,
xaxis=dict(title='Classe Prédite'),
yaxis=dict(title='Classe Réelle')
)
st.plotly_chart(fig)
def plot_CM_DenseNetFT():
confusion_lines = [
[285, 1, 2, 0],
[3,235 , 23, 0],
[5, 17, 232, 5],
[0, 0, 3, 265]
]
confusion_matrix = np.array(confusion_lines)
class_names = ['Covid', 'Lung opacity', 'Normal', 'Viral Pneumonia']
size = max(confusion_matrix.shape)
fig = ff.create_annotated_heatmap(z=confusion_matrix, x=class_names, y=class_names, colorscale='Cividis')
fig.update_layout(
width=size*140,
height=size*140,
xaxis=dict(title='Classe Prédite'),
yaxis=dict(title='Classe Réelle')
)
st.plotly_chart(fig)
def plot_CM_ENetB4():
confusion_lines = [
[282, 2, 3, 1],
[4, 244, 13, 0],
[2, 22, 220, 15],
[1, 0, 0, 267]
]
confusion_matrix = np.array(confusion_lines)
class_names = ['Covid', 'Lung opacity', 'Normal', 'Viral Pneumonia']
size = max(confusion_matrix.shape)
fig = ff.create_annotated_heatmap(z=confusion_matrix, x=class_names, y=class_names, colorscale='Cividis')
fig.update_layout(
width=size*140,
height=size*140,
xaxis=dict(title='Classe Prédite'),
yaxis=dict(title='Classe Réelle')
)
st.plotly_chart(fig)
# GRAD-CAM
def make_gradcam_heatmap(img_array, model, last_conv_layer_name, pred_index=None):
model_output = model.output if isinstance(model.output, list) else [model.output]
grad_model = tf.keras.models.Model(
inputs=model.inputs,
outputs=[model.get_layer(last_conv_layer_name).output] + model_output
)
with tf.GradientTape() as tape:
last_conv_layer_output, preds = grad_model(img_array)
if pred_index is None:
pred_index = tf.argmax(preds[0])
class_channel = preds[:, pred_index]
grads = tape.gradient(class_channel, last_conv_layer_output)
pooled_grads = tf.reduce_mean(grads, axis=(0, 1, 2))
last_conv_layer_output = last_conv_layer_output[0]
heatmap = last_conv_layer_output @ pooled_grads[..., tf.newaxis]
heatmap = tf.squeeze(heatmap)
heatmap = tf.maximum(heatmap, 0) / tf.math.reduce_max(heatmap)
return heatmap.numpy()
def save_and_display_gradcam(img, heatmap, alpha=0.4):
heatmap = np.uint8(255 * heatmap)
jet = plt.cm.jet
jet_colors = jet(np.arange(256))[:, :3]
jet_heatmap = jet_colors[heatmap]
jet_heatmap = array_to_img(jet_heatmap)
jet_heatmap = jet_heatmap.resize((img.shape[1], img.shape[0]))
jet_heatmap = img_to_array(jet_heatmap)
superimposed_img = jet_heatmap * alpha + img
superimposed_img = array_to_img(superimposed_img)
return superimposed_img