Spaces:
Runtime error
Runtime error
# -*- coding: utf-8 -*- | |
""" | |
Created on Fri Dec 23 09:25:09 2022 | |
@author: Usuario | |
""" | |
from keras.models import load_model | |
import tensorflow as tf | |
from tensorflow.keras.utils import load_img, img_to_array, array_to_img | |
from keras.preprocessing.image import ImageDataGenerator | |
from keras.applications.vgg19 import preprocess_input, decode_predictions | |
import matplotlib.pyplot as plt | |
import numpy as np | |
from IPython.display import Image, display | |
import matplotlib.cm as cm | |
import streamlit as st | |
from tensorflow.keras.utils import load_img | |
import pandas as pd | |
import matplotlib.pyplot as plt | |
import numpy as np | |
import streamlit_toggle as tog | |
import explicability as ex | |
from bokeh.plotting import figure | |
from bokeh.palettes import Category20c | |
from bokeh.plotting import figure, show | |
from bokeh.transform import cumsum | |
last_conv_layer_name = "block5_conv4" | |
def resultados(uploaded_image, model, size, label_names, labs, result_text): | |
#st.image(uploaded_image, caption='Celula Sanguinea') | |
image = load_img(uploaded_image, target_size = size) | |
img = np.array(image) | |
img = img / 255.0 | |
img = img.reshape(1, 224, 224, 3) | |
label = model.predict(img) | |
#st.text([np.argmax(label)]) | |
score = label[0] | |
#st.write(add_text_chart) | |
#st.write('La imagen a analizar es la mostrada a continuación. Se podrá analizar la probabilidad de pertenencia a cada tipo de célula sanguínea, así como observar el mapa de calor generado en el apartado de explicabilidad.') | |
col1, mid, col2 = st.columns([300,300,300]) | |
with mid: | |
st.image(uploaded_image, width=220, use_column_width=False) | |
#with col2: | |
placeholder = st.container() | |
tab1, tab2 = placeholder.tabs(labs)#(["Result", "Explicability"]) | |
with tab1: | |
#st.write('Aquí se muestra la probabilidad de la imagen seleccionada de pertenecer a cada clase de célula sanguínea según el modelo de Inteligencia Artificial entrenado.') | |
st.write(result_text[0]) | |
st.write(' ') | |
#Bokeh pie chart | |
pie = {label_names[0]: np.round(score[0]*100, 2), | |
label_names[1]: np.round(score[1]*100, 2), | |
label_names[2]: np.round(score[2]*100, 2), | |
label_names[3]: np.round(score[3]*100, 2), | |
label_names[4]: np.round(score[4]*100, 2), | |
label_names[5]: np.round(score[5]*100, 2), | |
label_names[6]: np.round(score[6]*100, 2), | |
label_names[7]: np.round(score[7]*100, 2)} | |
datita = pd.Series(pie).reset_index(name='value').rename(columns={'index': 'country'}) | |
datita['angle'] = datita['value']/datita['value'].sum() * 2*np.pi | |
datita['color'] = Category20c[len(datita)] | |
p = figure(height=350, title="", toolbar_location=None, | |
tools="hover", tooltips="@country: @value", x_range=(-0.5, 1.0)) | |
p.wedge(x=0, y=1, radius=0.4, | |
start_angle=cumsum('angle', include_zero=True), end_angle=cumsum('angle'), | |
line_color="white", fill_color='color', legend_field='country', source=datita) | |
st.bokeh_chart(p) | |
#===================== | |
col1, col2, col3, col4, = st.columns([250,250,250,250]) | |
col1.metric(label_names[0], str(np.round(score[0]*100, 2))+"%") | |
col1.metric(label_names[1], str(np.round(score[1]*100, 2))+"%") | |
col2.metric(label_names[2], str(np.round(score[2]*100, 2))+"%") | |
col2.metric(label_names[3], str(np.round(score[3]*100, 2))+"%") | |
col3.metric(label_names[4], str(np.round(score[4]*100, 2))+"%") | |
col3.metric(label_names[5], str(np.round(score[5]*100, 2))+"%") | |
col4.metric(label_names[6], str(np.round(score[6]*100, 2))+"%") | |
col4.metric(label_names[7], str(np.round(score[7]*100, 2))+"%") | |
#chart = pd.DataFrame(np.array(score)*100, label_names) | |
#st.bar_chart(chart, use_container_width=True ) | |
#p = figure(title = '', | |
# x_range = label_names) | |
#p.vbar(x = label_names, top = np.array(score)*100) | |
#st.bokeh_chart(p, use_container_width= True) | |
# fig, ax = plt.subplots() | |
# ax.bar(label_names, np.array(score)*100, color = 'red') | |
# st.pyplot(use_container_width = True) | |
with tab2: #Explicabilidad | |
#st.write('El mapa de calor generado con el algoritmo GRADCAM es el mostrado a continuación. En él se puede observar qué parte de la imagen de entrada ha sido la parte más relevante para el modelo de Inteligencia Artificial en cuanto a clasificación se refiere.') | |
st.write(result_text[1]) | |
col3, col4, col5 = st.columns([300,300,300]) | |
with col4: | |
img_array = preprocess_input(ex.get_img_array(uploaded_image, size)) | |
model.layers[-1].activation = None | |
heatmap = ex.make_gradcam_heatmap(img_array, model, last_conv_layer_name) | |
st.image(ex.save_and_display_gradcam(uploaded_image, heatmap), use_column_width=True) | |
def idioma(): | |
idiomita = tog.st_toggle_switch(label = "", | |
key = 'he', | |
default_value = True, | |
label_after = False, | |
inactive_color="#ffffff", | |
active_color="#ffffff", | |
track_color="#18202b" | |
) | |
return idiomita | |
def change_title(idioma): | |
if idioma == 1: | |
title = st.title('Peripheral blood cells classification') | |
#lab = ["Result", "Explicability"] | |
else: | |
title = st.title('Clasificación de imágenes de células sanguíneas periféricas ') | |
#lab = ["neeee", "bruuu"] | |
return title | |
def change_labels(idioma): | |
if idioma == 1: | |
labs = ["📈 Result", "📝 Explicability"] | |
else: | |
labs = ["📈 Resultados", "📝 Explicabilidad"] | |
return labs | |
def add_bg_from_url(): | |
st.markdown( | |
""" | |
<style> | |
.stApp { | |
background-image: linear-gradient(gray,white); | |
background-size: cover | |
} | |
</style> | |
""", | |
unsafe_allow_html=True | |
) | |
def button_image(): | |
st.markdown( | |
f""" | |
<button type="submit"> | |
<img src="https://i.ibb.co/CW5Wvry/buttonpng.png" alt="buttonpng" border="0" /> | |
</button> | |
""") | |
def additional_text_chart(idioma): | |
if idioma == 1: | |
text = 'The following image is going to be analysed. In **Results** you can observe the probability that this cell has to belong to a determined blood cell type, and the color map in **Explicability**' | |
else: | |
text = 'La imagen a analizar es la mostrada a continuación. Se podrá analizar la probabilidad de pertenencia a cada tipo de célula sanguínea, así como observar el mapa de calor generado en el apartado de explicabilidad.' | |
return text | |
def result_text(idioma): | |
if idioma == 1: | |
textito_res = 'Here appears the probability of the input image to belong to each blood cell type depending of our IA trained model.' | |
textito_exp = 'The color map was generated with GRADCAM algorithm. Here you can observe which part of the input image has been the most relevant part for the IA model in terms of classification.' | |
textito = [textito_res, textito_exp] | |
else: | |
textito_res = 'Aquí se muestra la probabilidad de la imagen seleccionada de pertenecer a cada clase de célula sanguínea según el modelo de Inteligencia Artificial entrenado.' | |
textito_exp = 'El mapa de calor generado con el algoritmo GRADCAM es el mostrado a continuación. En él se puede observar qué parte de la imagen de entrada ha sido la parte más relevante para el modelo de Inteligencia Artificial en cuanto a clasificación se refiere.' | |
textito = [textito_res, textito_exp] | |
return textito | |
def botoncitos(idioma): | |
if idioma == 1: | |
label_pj = 'About the project 📕' | |
label_us = ' About us 🧝♂️ ' | |
labelcillos = [label_pj, label_us] | |
else: | |
label_pj = 'Sobre el proyecto 📕' | |
label_us = ' Sobre nosotros 🧝♂️ ' | |
labelcillos = [label_pj, label_us] | |
return labelcillos |